OBSTACLE AWARENESS AND COLLISION AVOIDANCE RADAR SENSOR SYSTEM FOR LOW-ALTITUDE FLYING SMART UAV

OBSTACLE AWARENESS AND COLLISION AVOIDANCE RADAR SENSOR SYSTEM FOR LOW-ALTITUDE FLYING SMART UAV Young K Kwag and Jung W Kang, Avionics Dept. AERC, Hanhk Aviation University, Seoul, Korea Abstract In this paper, the critical requirement for obstacle awareness and avoidance is assessed with the compliance of the equivalent level of safety regulation, and then the collision avoidance sensor system is presented with the key design parameters for the requirement of the smart unmanned aerial vehicle in low-altitude flight. Based on the assessment of various sensors, small-sized radar sensor is selected for the suitable candidate due to the real-time range and range-rate acquisition capability of the stationary and moving aircraft even under all-weather environments. Through the performance analysis for the system requirement, the conceptual design result of radar sensor model is proposed with the range detection probability and collision avoidance performance in a typical flight environment. Introduction Recently, Unmanned Aerial Vehicle (UAV) has been drawing a great attraction for the applications to both civil and military mission without risk in air safety for the pilot flying in lowaltitude and/or in dangerous battlefield environment. Due to the inherent nature of the low flying vehicle, obstacle awareness is a fundamental requirement to avoid the collision against stationary and/or moving target obstacles along the flight path. Also it is noted that UAV should secure the equivalent level of safety comparable with manned aircraft in order to fly in civil and military airspace. Thus the collision avoidance system should be considered as a part of the navigation system in an unmanned vehicle. The obstacle awareness and collision avoidance is being acknowledged as the most important issues in the field of unmanned vehicle. An international standard regulation for the sense-and-avoid of UAV is being studied in Europe and USA, and the related technology has not been matured yet for the unmanned systems application. Recent technology advances in obstacle detection using optical imaging sensor for aircraft [I] and laser radar for helicopter [2] have been addressed in the articles, but the radar sensor technology for obstacle awareness and collision avoidance is being under development [3]and has recently been demonstrated for the manned helicopter and unmanned vehicle [4]. Most of current collision avoidance system of the manned aircraft employs Traffic Alert and Collision Avoidance System (TCAS) with standard protocol, which measures relative range and range rate of intruding aircraft using interrogation method and provide the decision information to the pilot for the flight control autonomously [SI. ADS-B (Automatic Dependent Surveillance-Broadcast) is being considered for the future collision avoidance system, which broadcast own GPS-based flight position information to the pass-by aircrafts in order to secure the sufficient distance around own aircraft [6]. But many aerial vehicles, especially MAVs and UAVs can not afford to cany too many avionic aids due to the weight, volume, power, and even cost constraints. Thus, collision avoidance strategies require minimal avionic aids and effective performance with small-size and light-weight for the pilot and even for automating the task of collision avoidance in low altitude pilotless aerial vehicles. In addition, collision avoidance system should autonomously sense the obstacles along the flight path and measure the range and range-rate in real-time under the all-weather environments. The collision awareness and avoidance system is under feasibility study for the Korea Smart UAV project, whose objective is to develop a smart UAV capable of high speed cruise and vertical takeofulanding (VTOL) by integrating smart technologies over the years, funded by the Ministry of Science and Technology. The basic system requirement includes max. speed of SOOKm/h, altitude of 3 Km, endurance time of 3-5 hours, and obstacle detection range of more than 5 Km. In this paper, the characteristics of candidate sensor for collision avoidance are compared and the requirement for the radar sensor is assessed in terms 0-7803-8539-X/04/$20.00 2004 IEEE 12.D.2-1

of the system requirement and air safety regulation. Radar sensor-based obstacle awareness and collision avoidance system is proposed with the radar design parameters and avoidance criteria. Finally, the millimeter radar detection performance and the collision avoidance algorithm are presented for the application to the unmanned vehicles in the typical flight environments Characteristics of Collision Avoidance Sensor System In general, the collision avoidance system for the manned and/or unmanned vehicles can be divided into cooperative system and noncooperative system. The examples of the cooperative system include TCAS, ACAS, which currently is equipped with the general manned aircraft, and ADS-B is not dedicated for the collision avoidance system, but is under consideration for the future application. Noncooperative system employs the sensors such as electro-optical sensor and electromagnetic radar sensor to measure the avoidance information of the obstacles using active or passive metbod. Active sensor emits energy to obstacles and receives the reflecting energy, while passive sensor only receives energy radiated from obstacles. Depending on the type of applicable sensor, there are a number of advantages and disadvantages in terms of operating frequency, average power consumption, detection range, component size, angular resolution, atmospheric effects, clutter characteristics, and electromagnetic interference immunity. Due to the limited Capability of measuring the real-time range, the passive sensor is not considered for the candidate comparison. The frequency band of the microwave radar sensor is relatively low, so it does not provide the sufficient angular resolution, but is insensitive to the weather situation. On the other hand, the millimeter wave radar provides the advantages of the small-size, fine angular resolution, but it is limited to detection range and is sensitive to weather environment. Angular resolution may be improved by increasing the operating frequency, but is physically dependent on the aperture size of the antenna. Table 1 summarizes the characteristics of active sensors in terms of the obstacle detection capability. Condition Table 1. Sensor Characteristic Micro wave MMW Mode Active Active Wavelength cm cm Average same range) Range/Range -rate value Detection Range Component Angular resolution Thermal Imager Active lun I power (in I High I Medium I Low I effects Not - Available Available available Unlimited lokm 10-15Km Large Small Small Large Small Small Degraded As a cooperative collision avoidance system, the TCAS provide a set of electronic eyes so the pilot can "see" the traffic situation in the vicinity of the aircraft. TCAS displays to show the pilot the relative positions and velocities of aircraft up to 40 miles away. The instrument sounds an alarm when it determines that another aircraft will pass too closely to the subject aircraft. There are two different versions: TCAS I, indicates the hearing and relative altitude of all aircraft within a selected range in IO to 20 miles. When pilots receive a Traffic Advisory (TA), they must visually identify the intruding aircraft and may alter their plane's altitude by up to 300 feet. TCAS I1 also provides pilots with resolution advisories (RA's) when needed. The system determines the course of each aircraft; climbing, descending, or flying straight and level. TCAS II then issues an RA advising the pilots to execute an evasive maneuver necessary to avoid the other aircraft, such as "Climb" or "Descend." Standards for ADS-B are currently being developed jointly by the FAA and industry through RTCA Inc. The concept is that aircraft or obstacle broadcast a message on a regular basis, which includes their position such as latitude, longitude and altitude, velocity, and possibly other information. Other aircraft can receive this information for use in a wide variety of applications 12.D.2-2

such as collision avoidance. These position reports are based on accurate navigation systems, such as GPS satellite navigation systems. The accuracy is unaffected by the range to the aircraft. With the radar, detecting aircraft velocity changes requires tracking the received data. Changes can only be detected over a period of several position updates. These improvements in surveillance accuracy can be used to support a wide variety of applications to collision avoidance in the airport and airspace for improving safety in the near future. Requirement Analysis of Collision Avoidance Sensor The smart UAV program requires the collision avoidance capability to automatically sense and avoid the stationary obstacles and/or non-stationary moving objects along the flight path in the relatively low-altitude flying and rapid maneuvering environment. It is essential to consider that the applicable collision avoidance sensor should be capable of providing the real-time position data measurement and accommodating the given payload avionics constraints such as weight, volume, and power. Based on the system requirement, the key design parameters can be extended as follows: System Requirement Maximum speed of UAV is 500 Kmh without payload and 440 K with payload. The surveillance mission payload is limited to less than 40 Kg, and the separated collision avoidance payload is reserved for less than 25 Kg. Collision Avoidance Mode There require three critical modes of operation to decide the status of the collision risk search, awareness, and avoidance modes. In a navigation mode, the system searches and monitors the obstacles within a certain scan volume along the flight path. Once the obstacles are detected from the sensor in a 23 sec of time-to-collision, the awareness mode is activated by tracking the position of the designated object. In an avoidance mode, the system initiates to maneuver and autonomously tum the vehicle in order to avoid the dangerous obstacles. At the closing speed of loookm/h, the sensor should detect the obstacle in a minimum distance of 6.4 Km. The types of obstacles include stationary objects such as manmade buildings, towers, power line and natural trees, hills, and mountains. The non-stationary objects are the closing and opening aircraft, helicopters, UAV, and other flying vehicles. Sensor Requirement The collision avoidance sensor selection should be considered in several important points of view: real-time measurement capability, operational environment, payload constraints, and the air safety regulation. The most important requirement of the obstacle detection sensor for collision avoidance is the capabilities of real-time measurement of relative range, range-rate, and bearing in azimuth and/or elevation. The operational environment of the vehicles is to be considered as the sensor selection criteria, which include the search and scan capability, all weather and day or night operational capability, vehicle maneuverability, endurance time in air, and ECCM capability. The most critical constrains for small unmanned vehicles are the payload requirements. The payload weight requires less than 25 Kg, and the volume limit requires 620 (width) x 420 (height) x 300 (depth) for the space of the collision avoidance equipment in the vehicle. A low power and long life time reliability are to be required as well. Final consideration is to meet the air safety requirement flying in airspace as directed by FAA or intemational standard. Unfortunately, the standard for sense-and-avoid requirement for UAV air safety has not been available, but at moment, as a upper bound of the regulation, the minimum requirement is to comply with the equivalent level of safety (ELOS) for see-andavoid for manned aircraft pilot by the FAA regulation. Table 2 summarizes the key regulation for collision avoidance of the manned pilot. From the FAA and EA0 regulation shown in Table 2, the search volume of the sensor must be +/- 60 degree in azimuth, and +/- 10 degree in elevation respectively, and minimum distance for collision safety must be more than 500ft [7]. It is to be secured for the pilot reaction time (PRT) of 11 sec within total time-to-collision of 23Sec to avoid collision safely, but for the unmanned vehicle, it is 12.D.2-3

5 reasonable that the reaction time could be significantly reduced by the maneuvering command of DFCC. Table 2. "See-and-Avoid" ELOS erformance ammeter MissedDistance ~ 00 feet 1 Field of Regard Search Volume -Azimuth,+I- 60" Collision Avoidance Radar Sensor OACAS Concept The conceptual design block diagram of the Obstacle Awareness and Collision Avoidance (OACAS) system shown in Figure 1 consists of CAS radar sensor and OCAS processor. TCAS and ADS-B are of option. I I I Through the initial phase of feasibility study on collision avoidance requirement, the noncooperative methods using active radar sensor is being selected for the primary collision avoidance sensor. There some trade-offs exist between microwave and millimeter radar sensor. Microwave radar sensor can satisfy the most of the requirements for sense-and-avoid mentioned above in detecting the range and range-rate, except for the limitation of payload accommodation due to the bulky sue and volume in a relatively long wavelength of microwave band. On the other hand, small-sized, light-weight millimeter wave radar can meet the equivalent level of requirement, even with the theoretically limited detection range of 10 Km. But this limit is within the collision avoidance requirement of detection range of 6.4 Km. From the trade-off assessment, the millimeter wave radar sensor is being selected for the primary collision avoidance sensor. However, the cooperative method such as TCAS and/or ADS-B could be considered to be an altemative option for the back-up system reliability as long as the payload accommodation margins are allowed. Figure 1. Concept of OACAS The CAS system consists of radar sensor (CAR) and Obstacle Collision Avoidance System (OCAS) processor. Utilizing the radar sensor data of range, azimuth or elevation, and velocity of the obstacles, the OCAS should decide the collision criteria and send the avoidance command to the DFCC (Digital Flight Control Computer), based on the time-to-collision criteria. The OCAS processor may be contained within the radar data processor or may be included in the DFCC. The CAS radar constitutes the antenna, transmitter and receiver, signal processor, and data processor, which will be described in detail. The OCAS processor computes the range rate to generate the collision avoidance command using the radar data. The OCAS may be contained within the radar data processor or may be included in the DFCC. Depending on the closing or opening speed of moving obstacles from the radar, the minimum required time-to-collision can be summarized in Table 3. In this case the UAV speed is assumed to be 440Km/h with payload. 12.D.2-4

Table 3. Time- to-collision Criteria the collision avoidance as well as surveillance using single radar hardware. The design specification can be traded-off on each parameter which is selected for the proposed radar system design. Table 4 shows the key radar system design-parameters to be decided Table 4. Radar Design Trade-off Parameters The CAS should comply with the volume and weight requirements for accommodating to the smart UAV system. As shown in the UAV system configuration, the CAS is installed in the nose part in front of the vehicle in order to efficiently sense and avoid the obstacles by searching the given scan volume. As the CAS accommodation shown in Figure 2, the volume size is 62 cm (width) x 42 cm (height) x 30 cm (depth), and the volume space of antenna is 30 cm in diameter. The weight of payload for CAS is limited to 25 Kg. N....I / I *,1L Width Antenna - I Elevation or Wider I 38 & bainlobel or More I Gain I -32 &'(sidelobe) or Less I Sidelobe Prob. Of 1 90% for SW2. RCS 1 m2. I Detection' I Pfa=lO-6 / Figure 2. CAS Accommodation Requirement CAS Radar Design Parameter The collision awareness and avoidance radar should detect and autonomously avoid the obstacles within the search volume during the flight in accordance with the CAS requirements above. The collision avoidance radar (CAR) system may be designed and implemented in a various ways for meeting the collision avoidance performance requirements. The dual modes of CAR and SAR (Synthetic Aperture Radar) may be considerable for Figure 3. Scan Volume of CAS radar \ 12.D.2-5

~ ~ CAS Radar Design Model A CAS radar design model for collision avoidance is shown in Figure 4. The radar system model consists of antenna, transmitter, receiver, signal processor, and system control computer. In addition the OCAS processor and the obstacle clutter map (C-Map) may be included for the OACAS mission. The key functional sub-system is to be specified as followings items: Antenna - Antenna Type, Servo Control, Gimbal - Scan Rate, Beam Searching Mechanism - Beam Width, Gain and Sidelobe Level Transmitter - Power Module (MPM, solid-state TRM, Magnetron, TWT) - Transmitting Power, Transmitting Frequency, COHO, STALO - PRF, Pulse width, Waveform Generation, Receiver - Receiver Noise Figure, I/Q Detector or Synch Detector - Phase Noise, Coherent Receiver or Coherent-on-Receiver Radar Signal Processor - A/D converter, MTI, CFAR, or DFB(FlT) - Pulse Compression, DSP, Clutter Filter, Clutter Map System Control Computer - Target Tracking, System Control Processor - Target Display, Radar Data Extraction Scan Rate Antenna Beam Width Antenna Gain RCS Prob. of False Alarm Prob. of Detection [ DFCC Figure 4. Collision Avoidance Radar The typical radar design parameters are listed in Table 5. Table 5. Typical Radar Sensor Model 150 deg/sec 2.5 deg 38 db 2-30 dbsm 10e-6 9O%(SWmode2 & RCS I2dB) For the given radar design parameter, the detection range performances is shown in Figure 5 in terms of SNR and number of pulses during the dwell time for the integration of 16 pulses. At the maximum closing speed of loookm and the RCS of 2 db, which is lower bound of small obstacle, the SNR is required more than 7 db for the detection range of 6.4 Km, which meets the given requirement. System Interface - Radar Interface to/from FCC (Flight Control Computer) - Electrical Interface, Mechanical Interface 12.D.2-6

Radar Detection and Collision Avoidance Performance The collision avoidance problem using radar sensor may be separately considered in two cases. One is the awareness problem, which can be represented by the probability of detecting the obstacles which are statistically different RCS model in the unmanned vehicle environments. The other is the avoidance problem which can be represented by the performance of the collision avoidance algorithm based on the given radar range, range-rate, and bearing information in the various flight scenarios. The collision avoidance algorithm also affects on the available data in real-time as well as the accuracy or errors of the range, range rate, and bearing information. Thus the awareness problem for detecting the accurate radar information from the radar sensor is first resolved before the avoidance mode is activated to turn the vehicle in a sufficiently safe distance. In the awareness mode, the OCAS processor can just provide the moving status of all the obstacles along the flight path by tracking a number of obstacles, and in the avoidance mode, the OCAS processor should generate the command data of which way is the best flight path in order to evade the intruding obstacles. In order to provide the accurate flight maneuvering data, the probability of obstacle detection is more significant issues in the awareness mode. Obstacle Detection Problem The radar obstacle can be represented by the stationary objects (buildings, towers, trees, mountains) and the non-stationary objects (aircraft, helicopters, flying vehicles). The radar cross section (RCS) of the object is the function of the reflectivity coefficient and the size of the reflected surface. It also varies with the aspect angle, wave length, and multi-path effects. Due to the difficulties associated with the exact RCS prediction even for simple shape objects, approximate method become the viable alternative. A complex target RCS can be modeled as a group of individual scattering centers distributed over the objects. Stationary obstacles can be modeled as static RCS, but as vehicle moves, the backscattered RCS may vary with the aspect angles. In most practical situation, there is relative motion between the radar and observed objects. Thus, the RCS measured by the radar fluctuates over a period of time as a hnction of frequency and the object aspect angle. This dynamic RCS may be fluctuates in amplitude (scintillation) and/or in phase (glint). For most cases, the glint can not be of major concem, but the cases where high precision and accuracy required, glint can be a detrimental. Scintillation can vary slowly or rapidly depending on the obstacle size, shape, dynamics, and its relative motion with respect to the radar. Due to the RCS changes as random process, the RCS scintillation model is used as a statistical model. The Swerling model [SI is used for this simulation [9]. For given radar model parameter and the probability of false alarm, the probability of detection is compared with the SNR of the radar pulses integrated. Figure 6 and Figure 7 show the probability of detection for the obstacle of Swerling model I and 11, which are relatively slow, and fast moving aircraft respectively, at the fixed false alarm probability of IO". Figure 8 and Figure 9 also show the probability of detection for the obstacle of Swerling model I11 and N, which are relatively slow, and fast moving aircraft respectively, at the fixed false alarm probability of IO". All these results show that 90% of probability of detection can be maintained when the 16 number ofpulse are integrated for the RCS of 2dB of intruding aircraft objects. 12.D.2-7

0.q 0.1 1 oi :., -L.d -10 d 0 J 10 IS 20 21 so SNR.dB Figure 6. Prob. of Detection for SW.l(Pfa-10-6) Figure 7. Prob. of Detection for SW.2(Pfa=10? I / 0.q. ' 0L-I, i -10 5 0 I 10 15 20 25 30 SNR-W Figure 8. Prob. of Detection for SW.3(Pfa=10d) i 1 Figure 9. Prob. of Detection for SW.4(Pfa=104) Collision Avoidance Problem The collision avoidance problem is to determine the possibility of collision, and to provide the maneuvering command to the vehicle in the complicated obstacle environments. There are three modes of operations to be conceived: search mode (detect), awareness mode (detect and track), and avoidance mode (maneuvering). The procedure of the collision avoidance is shown in Figure 10. In search mode from the normal navigation state, the UAV flies to the pre-programmed destination by navigation system, and CAS radar searches the obstacles within the scan volume along the flight path, and detects the obstacles and sees the potential obstacle coming close to own vehicle. Once any objects comes close within the time-to-collision of 23sec, the search modes is transferred to the awareness mode, which can now react the obstacle by tracking the moving obstacle by updating the vector of tbe object every scan rate. If the time-tocollision of any closing objects is less than 11 sec during the track, then the avoidance mode is initiated to maneuver the vehicle. Otherwise, if the threat is removed, then the normal navigation state is returned while searching any threats. The collision avoidance performance primarily depends on the available radar measurement and its accuracy of the range, range-rate, and bearing in the various flight scenarios [lo, 111. 12.D.2-8

n"ll sec1 Reaction Time Navigation State 1 [Searching, Obstacle - Detecting Obstacle Range Velacity, Bealin~ mparilon to cnteri Tracking Obstacle Range velosny, 0ean"E Range Velocily [Vu. vi) eann Beating Reaction Bearing I Maneuvering for Avoidance i Command to DFCC Search ;:: Figure 10. Procedure of Collision Avoidance A typical flight scenario could be established to investigate the problems involved in the avoidance mode. The key radar parameters for modeling the flight situation include the velocity vector, and angular aspects of azimuth and elevation in the given distance. The first scenario could be the simple case where the moving obstacle is closing in the radial direction whose aspect angle is 0 degree. In this case, depending on the speed of the closing intruder, the time-to-collision is determined by computing the range rate, abd is compared with the criteria. The second scenario is the more general case where the moving obstacle is closing in the angle direction &om 0-90 degree. In this case the closing speed of obstacle is varied by aspect angle, so it is more sensitive in getting the stable range rate information: fd =/2 2vr case where, fd is Doppler shift, v, is closing speed, h is wavelength, 8 is aspect angle. If the aspect angle tl is go", it will not be able to measure time-tocollision because fd is zero. In case where the aspect angle ranges between Oo and 90, the timeto-collision is calculated by utilizing closing speed and range information. The third scenario is the case where the obstacle vehicle flies to the opening direction, but own vehicle moves fast than the obstacle speed. In this case, there is a collision possibility in only case where the speed of own vehicle is faster than that of the opening obstacle. The fourth scenario is the case where the stationary obstacles such as tower or hill are within the threat distance. In this case, the collision avoidance is possible by using range, bearing, and clutter map information. The more practical situation should be considered for the various flight tests. The collision avoidance algorithm is also evaluating using the open algorithm for the possibility of radar sensor applications. Concluding Remark Due to the inherent nature of the low flying vehicle, obstacle detection along the flight path is a fundamental requirement to avoid the collision against stationary andor moving target obstacles. In this paper, the critical sensor characteristic for obstacle awareness and avoidance is assessed with the compliance of the equivalent level of safety regulation. The two collision avoidance systems are compared cooperative and non-cooperative system. Based on the assessment of the obstacle awareness and collision avoidance sensor, the small-sized, light-weighted radar sensor is proposed for the suitable candidate in meeting with the system requirement as well as operational requirement of smart unmanned vehicle. The conceptual radar design result is also presented with the design parameters, and the radar detection and avoidance procedure are presented with the probability of obstacle detection and the avoidance scenarios. It is desirable to approach to the synthetic method by combining the advantages of the noncooperative radar sensor as well as cooperative TCAS and ADS-B to efficiently increase the probability of collision avoidance for the nonstationary moving aircraft as well as stationary obstacles in the low-flying UAV. Future work in this study would address the implementation of the radar sensor and the avoidance algorithm evolution for the miniature-radar sensor for efficient collision avoidance system. 12.D.2-9

Acknowledgement This work was supported by the smart UAV Development Center, Korea Aerospace Research Institute, Daejeon, Korea. The authors wish to acknowledge Dr. Hyun C. Lee (Avionics Team) and Dr. Chul H. Lm (Director) of the smart UAV Development Center, Korea Aerospace Research Institute, for their supports and directions for this work. References [I] Tarak Gandhi and et al. 2001, Detection of obstacles in the Flight Path of an Aircraft, IEEE Tr. AES Volunme:39, No.1, Pages:l76-191. [2] Anthony Mangogno, and et. AI., 2001, Development of a Helicopter Obstacle Detection and Air Data, IEEE 20* Digital Avionics Conference, Volume:l, Pages 1-18, [3] B. Kumar and D. Ghose, 2001, Radar-Assisted Collision Avoidance /Guidance Strategy for Planar Flight, IEEE Tr. AES. Volume: 37, No.1, Pages 77-90. [4] Russel Wolfe, 2003, Non-Cooperative Collision Avoidance Flight Test Results & Analysis, European W S Tech 2003 Conference. [5] Williamson T., Spencer N.A., 1989, Development and operation of the Traffic Alert and Collision Avoidance System (TCAS), Proceedings of the IEEE,Volume: 77, Pages1735-1744. [6] Robert Holdsworth, Jim lambert, 2001, In-flight path planning Replacing Pure Collision Avoidance Using ADS-B, Aerospace and Electronic Systems IEEE, Volume: 16, Pages27-32 [7] Russel Wolfe, 2003, Is Meeting an Equivalent Level of Safety for See & Avoid Adequate for ROA Operation within Civil Airspace?, European W S Tech 2003 Conference [8] Swerling, P, 1997, Radar probability of detection for some additional fluctuating target cases, Aerospace and Electronic Systems, IEEE, Volume: 33, pp. 698-709 [9] Lee, A.; Mason, M., 2002, MATLAB simulation for computing probability of detection, Radar Conference, Proceedings of the IEEE, pp.478-483 [ 101 B. Ajith Kumar, D,Ghose, 2001, Radar- Assisted Collision Avoidance/Guidance Strategy for Planar Flight, Transactions on Aerospace and Electronic Systems IEEE, Volume. 37, pp. 77-90 [l 11 Barfield, F., 2000, Autonomous collision avoidance: the technical requirements, Proceedings of the IEEE, pp.808-813 12.D.2-10