A LASER RANGE-FINDER SCANNER SYSTEM FOR PRECISE MANEOUVER AND OBSTACLE AVOIDANCE IN MARITIME AND INLAND NAVIGATION A.R. Jiménez, R.Ceres and F. Seco Instituto de Automática Industrial - CSIC Ctra. Campo Real km 0.2 La Poveda 28500 Arganda del Rey Madrid (Spain) e-mail:arjimenez@iai.csic.es Abstract: A new maritime navigation system based on a laser range-finder scanner is described in this paper. Initially, the advantages of the Ladar system over the existing radar techniques are outlined. Then, the requirements of the Ladar regarding angles of exploration, measurement range and precision for different kinds of maritime operations are analysed. The principal components of this system: the laser range-finder, the scanning unit and the data processing and displaying unit, are described in detail giving their main characteristics. The Ladar pseudo-image formation process is described. These images are then processed for the extraction of features of interest, which are continuously tracked for obstacle avoidance and precise manoeuvre operations using Kalman filtering techniques. Finally, sets of test are performed in different scenarios: open-sea navigation including entering the port and pier, and river navigation approaching bridges and locks. Some problems are identified, related to the maximum range achievable, the density of Ladar image, and other minor aspects, which need more work to be solved, but in general the results are quite promising and encourage us to continue working in this direction. Key words: Ladar sensor, Maritime navigation, Tracking 1. INTRODUCTION Existing Radar navigation aids for maritime and in-land ships are recognized not to be too adequate for precise manoeuvring and short-range obstacle avoidance. This is mainly due to its limited minimum distance of operation, which is normally above 300 meters. There are some modern inland radars performing quite well at even shorter distances, but are not reliable enough for narrow rivers or canals. Therefore, radar is useless at very short-ranges around the radar antenna, and for precise manoeuvres the skipper must use its visual abilities to perform a slow and manually supervised navigation. Additionally, statistics say that almost half of the accidents produced even at good visibility conditions, could be avoided with the support of special navigation aids [1]. Apart from obstacle collision avoidance, there are other applications that are demanded by ship users and companies to increase the efficiency of traffic in precise operations [2], such as, entering/leaving the port and berthing/clearing the pier for the maritime operation, as well as, overtaking slower vessels, passing between the bridge columns, passing under a bridge with critical height, entering/leaving the lock chamber or coupling with other vessels (Fig. 1).
Figure 1. Some common ship operations: (left) entering the port and berthing the pier; (right) river navigation passing through narrow bridge columns 2. OBJECTIVES AND SYSTEM REQUIREMENTS The objective of this work is the development of an improved radar-like sensor operating at short ranges (0 to 1000 meters), valid in any adverse weather conditions (rain, fog, night, ), to be integrated and used as a complement to existing ship instruments (radar, ECDIS, speedlog, compass, etc). It should provide an important navigation help to allow more precise manoeuvres in less time, increasing the efficiency of water transport, and specially it should help to avoid collisions or accidents triggering alarms or even actuating automatically on skipper controls. Some of the requirements for the sensor are outlined in Table 1 for each of the different modes of operation of a ship. It s concluded that in many situations a 10 cm accuracy is desired and the range should reach 1000 m for the maritime field and 500 meters should be enough for in-land navigation (rivers and canals). It is also desirable to have a 180 degrees horizontal scan and a narrower vertical range of 90 degrees below the horizon. The update frequency needed in most cases is 1 Hz, but for special cases where faster update rates are needed (approximately 10 Hz for approaching the pier and the lock chamber) a prediction strategy based on previously made records will suffice. Finally, a data representation on an alphanumeric or graphical display is needed, together with a quantitative and symbolic representation of object of interest (wall of pier/lock, obstacle, approaching ship, etc) which should be identified and tracked giving information of distance, course and size by using kalman filtering techniques [3,4] for motion estimation of the target. 3. THE LADAR NAVIGATION SYSTEM The Ladar technology was the technique of choice to meet most of the requirements above stated. Time-of-flight Ladar can easily measure ranges above 300 meters (depends on surface reflectance and atmospheric conditions), and its accuracy is better than 10 centimetres. Using mirror-based scanners is it also possible to get a wide scanning angle both in horizontal and vertical directions (depending on mechanical assembly design). Also, due to the typical lowdivergence of laser beams, it is possible to get a very good lateral resolution, and therefore is possible, if ever needed, to acquire dense images with appearances similar to photographs (the
disadvantage is that total image completion requires several seconds). Summarizing, Ladar is the most appropriated technology nowadays to fulfil our requirements. Table 1. Sensor requirements in terms of distance range, accuracy, and horizontal and vertical scanning ranges, for each ship operation Ship Operation Distance range Horizontal bearing range Vertical bearing range Accuracy (m) ( ) ( ) (m) Maritime field Open sea navigation 1000 [-30,30] [0,-10] 10 Entering/Leaving the port 1000 [-90,90] [0,-45] 1 Berthing/Clearing the pier 100 [0,90] [0,-85] 0.1 Mooring to the buoy 300 [-90,90] [0,-45] 0.1 In-land navigation By-passing other vessel 500 [-30,30] [0,-30] 1 Overtaking other vessel 500 [-30,30] [0,-30] 1 Passing between columns 500 [-90,90] 0 1 Passing under the bridge 500 [-5,5] [5,-5] 0.1 Entering/Leaving the port 500 [-90,90] [0,-45] 1 Berthing/Clearing the pier 100 [0,90] [0,-85] 0.1 Locking 100 [0,90] [0,-85] 0.1 Coupling with other ship 100 [0,90] [0,-85] 0.1 The Ladar system developed consists basically of three main components: A laser rangefinder, a scanner and a processing unit with a man-machine interface (Fig. 2). It has been designed to allow easy integration with other complementary on-ship equipment (compasses, ECDIS, GPS, etc). The most important performance parameters of our prototype are listed below (table 2). The laser range-finder is based on a novel diode pumped Nd:YAG solid state laser (wavelength 1.064 microns) with a pulse repletion rate of 15 khz. The peak power of this laser is high enough (1 kw) to reach long distances (up to 1000 meters) on most natural reflecting surfaces (reflectance above 0.3). In spite to its high power, the total energy emitted is low enough (1 micro joules) to do not cause any injure on human eyes when the system is scanning and measuring distances on a region where some persons could be present (people on board a ship or close to the pier). The laser beam divergence is 1.5 mrad, which implies that permits a lateral resolution of 15 cm at 100 meters distance. The reflected energy coming back from the hit surface is sensed by an APD (Avalanche Photo Diode) capable of detecting a power of 53 nw which is the estimated received power at 1 km distance. Table 2. Ladar technical specifications Horizontal scan angle Horizontal scan speed Vertical scan angle Vertical scan speed Beam divergence (spot diameter at 100 m distance) Max measuring distance in good/poor visibility conditions Min. measuring distance Updating frequency of the display Protection class of the ladar head 80 degrees 5,4 lines(horizontal scans)/s 0 340 degrees 15 degrees/s 1,5 mrad (0.15 meters) 500 / 200 m 1 m 1 5 Hz IP64
Figure 2. Short-range Ladar system and its interface to additional navigation aids The scanner used to capture a low-density range image is based on a commercially available scanner from the company Riegl (LMS-Z210). This scanner allows a quick horizontal swept (5.4 lines/sec) based on a triangular rotating mirror. The vertical and slow swept is made at a maximum speed of 15 degrees per second. The scanner is normally placed at the bow of the ship in horizontal position (Fig. 3), therefore the images obtained consist of a few horizontal lines (1 to 10 lines) each one having approximately 1500 range values. Therefore, Ladar images have a very good lateral horizontal resolution (hits overlap slightly), but have poor vertical resolution and there are gaps between horizontal lines that could cause low profile objects not being detected. Fortunately, ship rolling helps spreading the horizontal lines vertically. The man-machine interface has a built-in data processing system that provides a graphical representation of captured raw Ladar information on top of an ECDIS map (Fig. 4). Additionally, it is displayed higher-level information, namely targets, extracted by an interpretation process based on tracking estimation techniques. Next section gives more detailed information on this topic. Figure 3. Ladar system on-board a vessel
4. OBSTACLE IDENTIFICATION AND TRACKING A radar-like representation of raw Ladar data is useful but difficult to interpret by a skipper. We have developed a software that is able to analyze the Ladar images in order to extract a more precise, and easy to interpret information to be finally used by a person or by an automatic alarm system, for example, triggering a signal when a collision risk is detected [3]. The main task of this software is the transformation of the raw data within Ladar frames into a list of targets, each one containing the estimated information: position, velocity, course, size and confidence about the estimation. This set of tracked targets is typically small, i.e. the algorithms have transformed raw data into manageable and significant information (Fig. 4, left). Firstly, the program reads, de-noises and analyses the frame looking for patterns of interest (i.e., edges, continuous profiles) using a combination of range and received laser amplitude. After pattern detection, the target tracking starts. This stage is an estimation process that uses the history of patterns to update a state model. This process is based on an individual Kalman filter [4] for each target that uses a constant velocity movement model. Finally, individual target are integrated by clustering using similarity measures to provide even a higher abstraction level. Figure 4. Man-Machine Interface that integrates ECDIS map with ship position and Ladar scans (black traces). (left) maritime; (right) in-land operation examples. 5. TEST AND RESULTS Tests were conducted in laboratory and on-board ships, both in river channels and in a maritime ferry [5]. Figure 4 (left) shows a snapshot of a travel between Salamis Island and Pireus harbour; two targets are detected: an approaching ship at 150 m distance and a vessel in front of our ship having the same course than ours. Figure 4 (right) shows a ship approaching a bridge in river Rhine; a bypassing ship is detected at the left hand side; the columns of bridges are easily detected. Other trials were performed at river Neckar, where ECDIS maps are not available (Fig. 5), under this conditions, narrow river and many locks,
the aid of a system like ours providing very short range information was quite valuable for navigation. Results are good and promising. It is possible to detect nearby boats and ships at distances up to 400 meters, the pier outline from Ladar data is clearly visible, shape of lock chambers are precisely sketched, the columns of bridges are reliably detected, and even the height of bridges can be measured with decimetre precision. On the contrary, some limitations have been detected, such us, the maximum achieved measuring range, which should be higher to allow the collision avoidance on rapid sea-going ships. Additionally, some noisy range measurements in images, due to reflections on foam-crested waves or debris, cause problems, so image-processing algorithms has to be improved to filter out outliers. Also, it will be of great help an active stabilization and levelling of the Ladar, or at least an image compensation based on roll and tilt sensors. Figure 5. (Left) Raw Ladar display approaching a lock chamber at river Neckar. (Right) a picture of the actual ship situation when the ladar scans where captured ACKNOWLEDGEMENTS Thanks to all partners that participated under the Shico Project for making possible the sensor integration and on-ship tests for the functional evaluation of Ladar interface and target tracking algorithms. REFERENCES [1] B. Zigic and J. Suomela, The short range navigation, positioning and obstacle avoidance system, Int. Conf. SURV V Surveillance, Pilot and Rescue Craft, Southampton, May, 2000 [2] VBD, Problem characterisation, Interim technical report on Task B.1, Shico project, Duisburg, June 1999 [3] IAI_CSIC, Pattern Identification and Tracking, Interim technical report on task E.2.3 and E.2.4, Shico project, Madrid, 2000 [4] Y. Bar-Shalom and X. Li, Estimation and Tracking Techniques, YBS, 1998 [5] HUT and VBD, Testing the pilot system - Inland waterways trial report, Interim technical report on task H, Shico project, June, 2001