LIDAR Odometry with Joint Geometric and Appearance Landmarks

lidar;odometry;maximum-likelihood estimation;continuous time;features;autonomous robotics

Skikos, Benjamin Anargyros Peregrineadvisor:Waslander, Steven ; affiliation1:Faculty of Engineering ; Waslander, Steven ;

University of Waterloo

关键词: features;    autonomous robotics;    odometry;    Master Thesis;    lidar;    continuous time;    maximum-likelihood estimation;   

Others  :  https://uwspace.uwaterloo.ca/bitstream/10012/14518/3/Skikos_Benjamin.pdf

瑞士|英语

来源: UWSPACE Waterloo Institutional Repository

【 摘 要 】

Odometry is the problem of estimating the motion of a moving platform relative to itsenvironment without measurements to a fixed reference point. This is a critical problemfor mobile robotics applications where measurements to a fixed reference point are notalways assured, such as self-driving cars operating in GPS-denied environments and withphysical landmarks potentially obscured by other traffic. Even when a fixed reference isavailable, motion estimates still improve position estimates by constraining the change inposition over short timescales.A fundamental limitation of odometry is that assuming some non-zero error, the pathproduced by integrating odometry will always diverge from the true path. The rate ofdivergence depends on the ego-motion and the environment. Aggressive motions suchas rapid rotation or high acceleration are likely to cause greater error because they aremore difficult to model compared to more sedate motions. Odometry that functions bycomparing consecutive sensor samples to determine motion can exhibit greater drift dueto high velocity because high velocity reduces the overlap between sensor samples. Lastly,unstructured portions of the scene may not contain useful information to fully constrainthe ego-motion. A good example is moving next to a flat featureless wall since observationsof that wall only constrain perpendicular motion.In mobile robotics, both camera and lidar are commonly used for odometry. Lidar is asensor technology that measures the time of flight of laser pulses to collect range-bearingsamples from the scene. Lidars are used instead of cameras for certain applications despitetheir relatively high cost because lidars are not affected by ambient lighting conditions;they do not suffer from glare, or have to trade-off motion blur and sensitivity in low-light conditions. An ancillary benefit of using lidar is that the direct range measurementcapability of lidar removes complexity from the odometry algorithm because the distanceof sampled points does not have to be estimated from multiple sensor measurements.Lidar odometry algorithms already exist yet there are opportunities for improvement.The intensity information collected by lidar commonly goes unused, with few of the top-performing lidar odometry algorithms on the Kitti odometry dataset leveraging it. As-suming that scenes lacking both geometric and appearance information are less likely thanthose lacking only geometric information, then a lidar odometry algorithm that leveragesboth types of information will be more robust than an algorithm that relies only on onetype, all other things being equal. Robustness is desirable because it makes performancepredictable.The main contribution in this thesis is a lidar odometry algorithm that uses bothappearance-based intensity landmarks as well as geometric landmarks to model the scene.The intensity landmarks are gradient edges that model features in the environment such aslane markings while the geometric features are geometric edges and planes. Feature pointsare selected from multiple lidar scans and matched to the landmarks. The landmarksand the trajectory are jointly optimized within a sliding window filter. The addition ofintensity landmarks improves performance in most cases compared to using geometric onlylandmarks.Evaluation is carried out using the Kitti odometry metric on the Kitti odometry datasetas well as a dataset collected in Waterloo Ontario. Average drift on the Kitti trainingset was 4.6-9.6 deg/km and 0.71%-2.8% translation drift. Average drift on the Waterloodataset was 3.1-3.4 deg/km and 1.1%-1.2% translation drift. Compared to a baseline con-figuration using only geometric landmarks, adding appearance-based landmarks produceda slight performance improvement on average across the Kitti subsequences as well as theWaterloo dataset, with up to a 13% reduction on translation error and a 6% reductionof rotation error for specific subsequences. However, there are also subsequences that ex-hibited worse performance with the addition of appearance-based features. In conclusion,the evaluation shows that the addition of appearance information can lead to better per-formance and that the method presented would benefit from additional work on furtherdeveloping the scene appearance model and how it is applied.