specifications as these arise from the requirements and the applications coming mainly from the automotive industry.

Transcription

1 issue 2 December 2010 Editorial Welcome to the MiniFaros EC funded project second newsletter. MiniFaros completed its first year and within this time the first outcomes have been made available. Within the year the partners have met regularly to discuss the work completed, fine-tune undergoing research and deliverables and plan ahead the upcoming tasks. The outcomes of this year comprise the definition of user needs and requirements and the laserscanner specifications as these arise from the requirements and the applications coming mainly from the automotive industry. These and a number of articles about the state-of-theart in sensor fusion, the optics and mechanics, as well as information about the omnidirectional lens are presented. Valuable information about the project, its advancements and also interesting technological material can be found in our updated website ( Inside this issue: Editorial 1 Requirements & User Needs 1 MiniFaros Architecture Specifications 2 SoA in Sensor Fusion 3 Laserscanners and Optics 4 News and Events 6 MiniFaros Consortium 7 MiniFaros Requirements & User Needs Accident review and analysis from European accident statistics have been performed and show that pedestrian protection and precrash functions are the main scenarios that can be addressed by an ADAS application using data from the MiniFaros laser scanner. In total around 54% to 82% of all severe accidents for cars and trucks could be addressed by ADAS safety functions using a MiniFaros laser scanner. The laser scanner can also be used for a number of ADAS comfort functions as stop and go and parking assistance. In this sensor development project ADAS functions already developed for safety will be adapted to the Mini- Faros laser scanner. There have been several European projects aimed for the development of safety functions and as a first step within MiniFaros a survey of the state-of-theart ADAS functions has been performed, in order to define the requirements for the laserscanner correctly. The requirements are described in functional terms regarding range, field of view, accuracy etc. The state-of-the-art ADAS survey indicates that the general requirements of the laser scanner sensor would

2 be approximately: Range: 80 meters Range accuracy: 0.1 meters in near-field and 0.3 meters else Field of view: 250 degrees Angular accuracy: 0.25 degrees Update frequency: 25Hz Additionally more general requirements on the laser scanner are also specified since the laserscanner has to comply with automotive standards, so as to be fitted and operated in vehicles. Object recognition algorithms have to be developed for the laser scanner for the safety applications addressed by the sensor. This includes the development of enhanced object recognition and tracking algorithms and performing improved object classification in order to be able to decide the correct strategy for avoiding objects in the traffic environment. The laser scanner will be shown and demonstrated serving various safety applications in vehicle environment. For this reason both a car and a truck demonstrator will be used. MiniFaros Architecture Specifications Laser scanners are the predominant generic environment sensing technology. The specifications for the laserscanner arise from the requirements defined within Workpackage 3 intending on the definition of a low cost sensor enabling multiple safety applications. The proposed design and composition of the sensor can be seen in the figure. Its central functional component is a twosided MEMS mirror rotating at 1500Hz, deflecting the outgoing laser beam by 30. This deflection is just enough to allow an omni-directional lens to guide the beam into the defined direction. At the same time, a second omnidirectional lens will guide the reflected, returning laser pulse back onto the backside of the MEMS mirror, into the APD. An FPGA is used to analyze the received laser pulses and derive the measured distance, direction and further information. A connected DSP analyzes the raw measurements provided by the FPGA in order to perform low level pre-processing and handle the data communication with the sensor. In this case, Ethernet is available for data transfer and configuration while the CAN interface is not foreseen to provide scan data but accept vehicle state information for optimized scanner internal pre-processing. In order to demonstrate the MiniFaros capabilities, 6 safety applications are specified in order to address the most relevant accident scenarios for trucks and passenger cars. They cover three pedestrian related as well as three frontal crash scenarios. The related HMI concept is described briefly and will be detailed throughout the project, while the system architecture is defined in detail. The software components and modules, performing the data processing from tracking and classification to the risk assessment inside the applications are specified as well. Besides the in-vehicle system specifications, an infrastructure demonstration system is specified as well. In contrast to the invehicle system, there are no applications defined for the infrastructure system. Instead, information on road users in the sensors vicinity (e.g. an intersection) is transferred via wireless communication in order to show the capabilities of the sensor in terms of cooperativeness.

3 State of the Art in Sensor Fusion Sensing the environment plays a very important role in many intelligent vehicle applications. A single sensor is unable to overcome certain physical limitations, such as the limited range and field of view. As a consequence, different sensors have to be exploited and fused for broadening the perception area around the vehicle and increasing the reliability of the whole system in case of sensor failure. In general data fusion first appeared in the literature in the 1960s, as mathematical model for data manipulation. In 1986 the US Department of Defense established a functional model, the Joint Directors of Laboratories (JDL), which is the most prevalent in data fusion. A revised JDL Framework is used in automotive fusion. It is a multilevel model that divides data processing in the following levels: signal, object, situation and application. The communication among the above levels is supported by a data management system for storage and retrieval of pre-processed data. The state-of-the-art in the automotive field is the fusion of many heterogeneous onboard sensors, e.g. radars, laserscanners, cameras, which are installed all around the vehicle. The coverage areas of these sensors can be either complementary or overlapping. In addition to the above mentioned sensors, GPS devices, inertial sensors and map data coming from digital map databases are incorporated in the fusion process. In this way, the estimation of the position of the subject vehicle is improved and extended even in locations where the GPS signal is not available (e.g. tunnels, bridges). Recently, V2x (x can be either V for vehicle or I for infrastructure) communication is used for enhancing the perception of the environment, broadening the perception horizon of the driver even to some kilometers ahead in some situations. Dedicated Short Range Communications (DSRC), which operate at 5.9 GHz, and Car2Car- Communication Consortium (C2C-CC), which is working towards the standardization of wireless communication between vehicles and roadside units, are initiatives focusing on the exploitation of wireless communication in vehicular environments. Coming to a conclusion, enhanced signal processing techniques must be explored and developed together with new advanced sensors (incl. maps and wireless communication as virtual sensors) to have better results and more accurate perception of the environment. Fusion of data coming from different information sources provides significant advantages over single sensor systems such as higher accuracy, reliability, completeness and confidence. Laserscanners and Optics There are automotive Laserscanners already available in the market from suppliers like Denso, Continental or Hella. However, the costs of such system are about Euros. In addition the field of view and the performance is often quite limited. Thus, mostly a single application is addressed, which directly is related to rather high costs per application. Furthermore there is a limited weather robustness, which limits the availability and reliability. Currently the smallest commercially available Laserscanner is the URG- 04L manufactured by Hokuyo in Japan aimed to serve industrial applications. The dimensions of this Laserscanner are 70 mm by 50 mm by 50 mm and it is available for about 600,-$. The scan angle is 240 degree and the measurement range is up to 4 m at white targets. However, the environmental robustness of the Hokuyo Laserscanner is not sufficient for automotive applications. A totally new sensor type will be developed in the range of laser sensors within MiniFaros. The new sensor will push state of the art in on the sensor level in the following areas: Affordable, target price of about 40 Euros, A very small size 4x4x4 cm enabling the use in

4 various locations of vehicles, work machines, moving robots and road side poles, Generic sensor possible to serve 12 different applications for future ADAS-platforms; also applicable to other than road vehicles only as mentioned above Durable, reducing the moving parts compared to current Laserscanners. capable of producing 360 degree panoramic depth maps. Video surveillance, autonomous navigation, and site modelling are some of the applications which would benefit from such a sensor. The main drawback of the presented configuration is that the volume that the measurement system requires prohibits its use in automotive applications. The manufacturing of cheap omnidirectional sensors haven Olympus. There are some problems with this kind of configurations in the system; usually these systems have high F-number, which means that optic throughput is low. Secondly, these kinds of surface shapes must be manufactured accurately and thus quite expensive, and occupy a large volume. Furthermore, Yamazava et al. also introduced basically a similar mirror system already in 1990 s. Today, there are several different types of omnidirectional (usually catadioptric configuration) optics on the markets like the one of Versacorp. There the lens has a diameter of about 12 cm, while the height is 14 cm. Both of the dimensions are unacceptable for automobile applications, and the costs are high due to injection moulding process. It has been, moreover, proposed an omnidirectional vision system that would provide stereo images of the scene. The design comprises of a compact panoramic stereo camera which uses parabolic mirrors and is been proposed already ten years ago. Several different methods are presented in the paper, but still the volume required by the system is prohibitive in automobile applications. Corke et al. employ omnidirectional vision sensor on top of the planetary rover. The omnidirectional mirror design proposed is too large, and thus cannot be used in the proposed Laserscanner in automotive environment. Hrabar and Sukhatme introduced an omnidirectional lens in their paper Omnidirectional vision for autonomous helicopter. The configuration is similar to a product of Vstone, AC- COWLE, RemoteReality and Currently, scanning lidar devices often use dual or nodding rotating polygonal mirrors, dual galvanometric mirrors, or a two-axis actuated mirror. These lidar scanning systems are bulky servo motor driven mirrors. They typically exceed a weight of several kilos. Driven by an increasing interest in automotive projection displays (head-up, dashboard) and also in portable projection displays for cell-phones, digital cameras, PDAs, game consoles etc., silicon based scanning micromirrors have become more and more attractive within the last years. Besides display applications these MEMS scanning mirrors have been developed also for laser scanning microscopy, endoscopic imaging, optical crossconnects, barcode scanning and many other applications. MEMS scanning mirrors

5 usually are fabricated using Silicon micro technology. Several hundred to some thousand MEMS mirrors can be fabricated in parallel on a single 6 inch Silicon wafer. Thus very lowcost mass production costs can be achieved. Several different actuation concepts have been realised so far to drive MEMS scanning mirrors: Thermal actuation provides high forces and enables large deflection angles. But it is limited to actuator frequencies of typically well below 1 khz. The material combination can cause problems with respect to hermetic packaging. Piezoelectric actuation is fast and provides high forces. But due to very low stroke it needs complex meander beam structures to enable sufficiently large mirror deflection angles. High forces are only achieved by use of PZT (lead-zirconiumtitanate) for which a mature and reliable deposition and structuring process has not yet become available. Electromagnetic actuation offers high forces, high speeds and also enables large reflection angles. However, the drawback is that mounting of permanent magnets on chip level is required. So it is in no way compatible to hermetic wafer level packaging technologies. Because of the need for external permanent magnets the total volume of that type of actuator is larger than for MEMS scanning mirrors driven based on other driving mechanisms. The most frequently used concept for driving MEMS scanning micromirrors is electrostatic actuation. Electrostatic forces offer very high speeds and do not require other materials than the wafer material itself. So in general this concept is compatible to wafer level packaging technologies. But a drawback of this actuation concept is the relatively low electrostatic force in comparison to electromagnetic force. Compatibility of MEMS scanning mirrors to automotive requirements is an issue that needs special concern. Typically automotive sensors require stable and reliable longterm functionality over a temperature range of -40 degree to +85 degree Celsius. So, in order to avoid failure by condensing moisture or by contamination by particles, fluids or gases, hermetic packaging of MEMS devices is a prerequisite and thus has become the standard for MEMS sensors. However; for MEMS actuators hermetic packaging is not yet widely established. Due to a limited cavity depth such fabricated scanning mirrors have a typical diameter of only 1 to 2 mm. Today, there are a number of industrial MEMS scanning mirror solutions on the market, like Microvision Inc. (USA), Samsung (South Korea), LG (South Korea), MEMS OPTICAL INC. (USA), Nippon Signal (Japan) or Fraunhofer (Germany). The novel omnidirectional plastic optics which will be designed in MiniFaros will push the art in the following aspects: Very low manufacturing costs by plastic injection moulding, Many optical functions integrated in a single component reducing the number of parts Very small size of optics enables a miniature sensor The entire optics within one component gives rugged and stable construction One of the main challenges in this proposed project is the development of a lowcost compact two-axes MEMS scanning system that can replace the expensive conventional technology. Thus, a 2D-MEMS scanning mirror will be developed that provides the following: A large mirror diameter of about 7 mm,

6 High flatness, i.e. mirror deformation < lambda / 10, Large mechanical deflection angles of > +/- 15 degree, Circular scanning by realisation of two axes of identical resonant frequencies High shock and vibration resistance (lowest resonance > 1kHz, vertical overload stops), Conformity to automotive requirements: hermetically sealed, Operating temperature range from -40 to +85 C, ambient condensing moisture, Low-cost mass producible (wafer-level packaging), Low chip size (preferably enabled by electrostatic drives), Integrated position feedback signal generation (e.g. capacitive). News and Events Minifaros 2nd Quarterly Meeting Tampere, June 29-30, 2010 On June 29-30, 2010 we had the second quarterly meeting that took place at the VTT premises in Tampere, Finland. It was organized by Uula Kantojärvi. The work completed so far was reviewed and open issues related to WP1 - WP5 were discussed in detail. Minifaros 3rd Quarterly Meeting Santorini, September 28-29, 2010 Santorini, Greece held the third quarterly meeting on September 28-29, 2010 kindly hosted by ICCS. Angelos Amditis was the coordinator of the meeting. During the meeting open issues to already running WPs were handled and more emphasis was given on WP7, since the actual work needed further planning. Technical discussions were mainly about the receiver and TDC, electronics and MEMS of the novice laser scanner. Minifaros 4th Quarterly Meeting Waldkirch, December 9-10, 2010 The fourth quarterly meeting took place in Waldkirch, Germany on December 9-10, Axel Jahn and Kay Fürstenberg were the organizers and hosts in SICK's headquarters. The status of the pending deliverables and work packages was examined and discussed. Bilateral discussions basically had to do with mechanical and optical issues related to the upcoming Minifaros laser scanner. Testing phase of the scanner was thoroughly discussed: when it will be conducted and consequently the action list for the upcoming period was updated.

7 issue 2 December 2010