Baraja and Monash Motorsport: Low Voltage Integration and Software-Definable Scan Patterns

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 Baraja’s Spectrum-Scan™ LiDAR.

M12 Connectors with the Baraja Sensorhead and engine

The flexibility of the Baraja Spectrum-Scan™ LiDAR has meant that resolving implementation challenges and meeting various rules, packaging, and safety requirements has been a relatively seamless process. The inherent versatility of the system, alongside the continual support and advice provided by the Baraja team, has enabled us to accomplish key objectives across both hardware and software. In this edition of our Baraja and Monash Motorsport blog series, we will be taking a deeper dive into Low Voltage Integration and software-definable scan patterns. These two examples help to illustrate the interplay between the various components of the perception sub-system and highlight the ways in which Baraja has simplified the development of our autonomous systems pipeline.

A key deliverable that our team needed to meet was the simultaneous implementation of the Baraja Spectrum-Scan™ LiDAR system within both M19-D and M21 vehicle architectures. With respect to this challenge, the Baraja Spectrum-Scan™ LiDAR offers a ‘plug and play’ solution to interface between 12V and 24V systems. This allows us to test both on and off the track, facilitating easy integration between our two driverless vehicles. To our benefit, the complex Baraja Spectrum-Scan™ LiDAR is efficiently packaged into a small system weighing around 3.4kg. This package consists of both the Sensorhead and engine, with the ability to be mounted anywhere on the vehicle. We tested a range of Sensorhead mounting solutions to find the ideal mounting position for our perception system. This package also comes with a custom mount for the engine unit, with an integrated fan that ensures adequate cooling is provided to the engine when in use. 

The supplied wiring harness consists of optic Fibre cables for data transfer between the Sensorhead, engine, and ethernet to our computing units. The use of standard connectors such as barrel plugs, M12 ethernet, and Deutsch connectors allow for easy integration into our electrical system. The team at Baraja have also been extremely accommodating in the supply of a custom harness for systems that require more efficient wiring packages. In addition, their exceptional ongoing support through the entire integration process has provided us with training and walkthroughs in cleaning and maintenance of the optic fibre cables and Sensorhead. Not only is the Baraja Spectrum-Scan™ LiDAR versatile in implementation but also robust across a variety of conditions including variable weather, vehicle architecture, and software operations. 

M12 Connectors

Concurrently, the ability to utilize software-definable scan patterns is new to our software development team and something that was not possible without the cutting-edge Spectrum-Scan™ technology. With our previous LiDAR unit, the pattern in which the environment was scanned was a fixed property, defined by the orientation of physical lasers inside the sensor. This meant that regions of the sensor's vertical resolution would be wasted depending on the manner with which the sensor was mounted. This is due to some of the returns being lost into the sky, or into the vehicle itself, thus providing information that is not necessary for our perception algorithms on the vehicle. As a result, the customizability of the Baraja Spectrum-Scan™ LiDAR Beamsteering system makes it highly suitable for our Autonomous Vehicle. As a comparison, our previous LiDAR was able to provide 16 vertical layers, whereas the Baraja Spectrum-Scan™ LiDAR is able to provide up to 1000 layers.

Software-definable scan patterns control the manner in which the environment is scanned by the Sensorhead, with the ability to focus the scan on different regions of interest throughout the scene. The patterns can also have either a linear or non-linear distribution of points with respect to the Sensorhead. In the development of an AV, this was pivotal in the ability to reliably detect objects at both short and long-range. Being able to request a dense distribution of points at a certain angle with relation to the Sensorhead resulted in the number of returns on each object of interest in the environment being similar, despite their various distances from the vehicle.

The team at Baraja, by providing extensive documentation on their products, have made the implementation learning curve a much easier challenge to face. The available software-defined scan patterns, with visual representations of each, were provided to the AV development team before receiving our unit. This even allowed us to shortlist which patterns we were expecting to perform well before testing the system on track. This shortlist was subsequently validated after testing every available scan pattern on track and confirming the best-performing scan patterns.

Software-defined scan patterns are able to be changed on the fly

Changing between different Scan patterns is possible through a simple REST (Representational State Transfer) API provided by Baraja, with requested changes to the Scan pattern being fulfilled nearly instantly. This opens up the opportunity for Scan patterns to be changed while the AV is completing events on track. In a Formula SAE environment, depending on whether the vehicle is traveling in a straight section of the track or cornering, it may be beneficial to have an increased density of points on specific areas of the scene. During cornering maneuvers, it would be best to focus LiDAR returns closer to the vehicle, to better identify the immediate bounds of the track. When traveling at a higher speed on a straight section of the track, information about the environment far ahead of the AV is more important. This is as with the fixed timestep duration of the entire pipeline, more distance would be covered in a single timestep when the vehicle is traveling at a higher speed. Meaning that in order for the autonomous systems to make predictions and actions by estimating the same length of time into the future, the vehicle must make use of perception information about points further away in the environment. Likewise, the opposite is true when performing lower speed maneuvers. Therefore, by being able to change Scan patterns, both these goals can be achieved without compromise, which is a large competitive advantage that the Spectrum-Scan™ technology has over traditional LiDAR systems.

Across both hardware and software integration challenges, the team at Baraja have been with us every step of the way. Thanks to this continual support, our team is progressing smoothly in achieving our Autonomous Vehicle goals in the development of our 2021 Driverless-Electric Formula SAE car, M21.