Baraja and Monash Motorsport: An Introduction to the Baraja Spectrum-Scan™ LiDAR

We’re excited to have our hands on our own Baraja Spectrum-Scan™ LiDAR system

Monash Motorsport has been focused on autonomous vehicle development since 2017, with the constant iteration and evolution of perception, path-planning and actuation systems. As part of its 2021 vehicle, M21, the team is excited to tackle its most ambitious vehicle concept yet, a fully autonomous, electric race car based on an all new vehicle architecture. As part of this ongoing development, this blog series aims to document and share the team’s progress in implementing a brand new LiDAR solution with the Baraja Spectrum-Scan™ LiDAR.

By partnering with Baraja, Monash Motorsport aims to gain an advantage over the top Formula Student / Formula SAE driverless teams in the world by directly addressing one of the biggest bottlenecks currently present within autonomous racing - a vehicle’s perception range. The Baraja Spectrum-Scan™ LiDAR system is able to provide incomparable performance from a single sensor head, with a maximum scan range of 200+ meters at a high frequency.

As part of the Formula Student competition, one of our goals is to navigate an unknown racetrack layout, with boundaries demarcated using traffic cones of varying colours and heights. 

The measurements returned by a LiDAR, known as Pointclouds, are used to identify objects based on a number of factors:

• The height of returned points above a ground plane

• The number of points returned within an empirically-determined physical distance

• If points in a similar position are returned in successive scans

Pointcloud density, usually affected by the vertical resolution (how small the vertical angle is between individual scan returns) of a LiDAR, is currently the main limiting factor in identifying objects at distance. Although each individual return may travel a large distance, the range at which objects are able to be identified is lower. When travelling, distant objects become effectively invisible as a result of increasing vertical distance between returns.

Due to this effect, with our previous system using a more conventional 16 channel, 360° rotating LiDAR, we were only able to reliably see the traffic cones on-track at a distance of around 7-8m ahead of the racecar.  This is a consequence of setting a minimum of 2 points within a set distance to reduce false positive detections. This limited range reduced the speed at which we could safely navigate around the track. In the case that there was a sudden change in direction, the vehicle would have been unable to react quickly enough in its steering and braking response. By increasing the range at which we are able to see, our motion planning systems are able to better predict the path that should be followed, and what actions should be taken.

Using the Baraja Spectrum-Scan™ LiDAR, which has a much higher vertical resolution, we are able to increase the distance at which we see cones and while also being able to derive greater certainty about the detections that we make. Already, during preliminary on-track testing, we are able to make cone detections from 40 meters away, which is almost 5 times greater than with our previous system. Where previously only a handful of returns would come from each cone, we are now able to see hundreds of returns from the nearest landmarks which has greatly improved the safety of our vehicle. Due to this high number of returns, we are able to explore classifying the type of cone in which we are detecting, which was not possible with our previous system. The varying intensity values of the Pointcloud from cone to cone enable us to classify differing cone colour patterns. This will reduce the possibility of navigating outside of track boundaries or misidentifying the layout of a corner. Thus, the robustness of the pipeline can be seen to have improved significantly. This is something for which we previously used a pair of stereoscopic cameras, which is no longer necessary because of the reliability and perception range of the Baraja Spectrum-Scan™ LiDAR.

Within an autonomous racing environment, improving the quality of object detections also has other flow-on effects for vehicles. The state estimation and motion planning algorithms depend on having an accurate and representative map of the car’s environment. Using an approach called Kalman Filtering for our Simultaneous Localization and Mapping (SLAM) algorithms, we try to account for uncertainty in all measurements we take. With more regular and accurate detections from the LiDAR, we can achieve more confident estimates for where the car is positioned in relation to the track and its surroundings. This translates into better real-world performance and improved robustness, as the vehicle is able to more closely follow the optimal path generated by our solvers.

Another important consideration within the wider Formula SAE / Formula Student competition is device packaging. Mass and space are at a premium on a racecar, and thankfully the Baraja Spectrum-Scan™ LiDAR has been very simple to incorporate into our new and existing vehicles. The compact nature of the units, as well as the support provided by the Baraja team at the physical, electrical, and software integration levels has allowed us to focus our efforts on making the most of the other benefits that the system provides.

The Baraja Spectrum-Scan™ LiDAR has the potential to take Monash Motorsport’s autonomous racing capabilities to the next level, and we are very excited to have Baraja on board, and working with us along the way. Stay tuned for more updates as our team makes its way toward unveiling our new fully autonomous, electric race car, M21.

Testing out placement of the Baraja Spectrum-Scan™ LiDAR on M19-D