Student-developed Driver Assistance Controls Reveal Students’ Creativity and Skills

Just as we threw out a challenge to students at the University of New Mexico to build a better flight simulator, we recently challenged some University of Toronto graduate students to use the Quanser immersive 3D environment and hardware-in-the-loop (HIL) vehicle simulation to design and test advanced driver assistance algorithms.

We did so for two reasons. First, we believe that hands-on learning using real hardware and immersive visualizations allows students to test and refine their otherwise “perfect” theoretical solutions in a real-world or near real-world context. Second, the design and integration of driver assistance systems through massive sensor fusion has been identified by the Institute of Electrical and Electronics Engineers (IEEE) as one of the grand challenges for control. By making this challenge so timely and relevant, we gave the students valuable exposure to the kinds of engineering problems they might encounter in their future careers.

We began by providing students with the basic platform and walking them through some fundamental control system labs to get them familiar with the platform. Then we asked them to think up some driver assistance challenges of their own and apply the collective engineering skills they’ve learned to a creative and relevant project. We didn’t hand-hold; we tried to get the students to work through the research, design and development themselves.

The students implemented some truly creative systems while following the recommended Quanser method throughout their development lifecycle from preliminary mathematical modeling, through simulation, HIL testing, and final deployment. Here are some of the most noteworthy results.

Car Following and Obstacle Avoidance:

One team chose to develop a series of algorithms to replicate a particular driving challenge, namely tracking and systematically passing a target vehicle while simultaneously avoiding randomly placed obstacles. This objective introduced several interesting control challenges and approaches including hybrid control, artificial potential field obstacle avoidance, state-feedback control, and the control of a steered vehicle. 

Visualization provided by the immersive 3D environment gave students an intuitive sense of worked and what didn’t.  This facilitated their development of algorithms that allowed one car to track and pass a target vehicle, while avoiding randomly place obstacles.

Once again, the students followed a systematic design process which began with the development of a mathematical model and control design, continued with the validation of their algorithms in simulation, before implementation on the test platform with actual servomotors as HIL components. Overall the students gained experience working on algorithms and techniques that have the potential to revolutionize the transportation industry. The feedback we received from the students indicated their use of the visualization tool helped them implement their mathematical models and see what worked and what didn’t.

Forward and Reverse Path Tracking with a Front-Wheel Steered Bicycle Model:

This team decided to design and compare controls systems for autonomous forward and reverse driving. Their emphasis was on the development of an accurate non-linear vehicle model to replicate a bicycle tracking an arbitrary path. The students were able to successfully implement their algorithms, and show some very impressive results and performance. 

Students discovered that developing algorithms with dynamic simulation models and hardware in the loop components helped to better predict controller performance once implemented in reality.

More important than their results, however, were their experiences working through the design process itself. The students made several critical observations including the need for accurate dynamic models for preliminary simulations, and HIL components when designing and testing control systems. The students remarked in their final report that, “Developing with these additions (e.g. dynamics, HIL) in mind can help to reduce the time required to tune real controllers once implemented, and can help to better predict the performance of a controller once implemented in reality. This is important because a controller might show very promising performance in a kinematic simulation without HIL, but performs very poorly even when tuned, once implemented on a real system.”

Success in engineering is only achieved when a challenge is met in theory AND in practice. The algorithms being designed have to work in the real world. To that end, the more we can bring engineering labs and projects into the real world through hands-on experiments, visualization and hardware-in-the-loop testing, the richer and more industry-relevant that education will be. That has been and always will be Quanser’s main focus.