Gradient-free Multi-domain Optimization for Autonomous Systems

Abstract

Autonomous systems are composed of several subsystems such as mechanical, propulsion, perception, planning and control. These are traditionally designed separately which makes performance optimization of the integrated system a significant challenge. In this paper, we study the problem of using gradient-free optimization methods to jointly optimize the multiple domains of an autonomous system to find the set of optimal architectures for both hardware and software. We specifically perform multi-domain, multi-parameter optimization on an autonomous vehicle to find the best decision-making process, motion planning and control algorithms, and the physical parameters for autonomous racing. We detail the multi-domain optimization scheme, benchmark with different core components, and provide insights for generalization to new autonomous systems. In addition, this paper provides a benchmark of the performances of six different gradient-free optimizers in three different operating environments. Our approach is validated with a case study where we describe the autonomous vehicle system architecture, optimization methods, and finally, provide an argument on gradient-free optimization being a powerful choice to improve performance of autonomous systems in an integrated manner.

Safe Reinforcement Learning for High-Speed Autonomous Racing

“…This shows that the baseline agent relies on the vehicle drifting for much of the track, thus exploiting the simulation model. This behaviour has been seen in other learning approaches [29], [30] and is responsible for causing the low completion rates. Policies relying on high-slip angles in the simulator are not feasible for physical implementation since in reality tyre dynamics are non-linear and thus the policy learned in simulation differs from how the real-world vehicles perform.…”

Resilient Design of Product Service Systems with Automated Guided Vehicles

“…One obvious example would be the synchronization of multiple AGVs in a narrow corridor of a storage building; for this kind of limited space coordination, detailed knowledge of the current and actual velocities would be very helpful. Zheng et al proposed using certain optimization methods for the joint optimization of the multiple domains of an autonomous system for finding an optimal architecture for both hardware and software [61]. Several researchers have proposed multi-domain co-simulation methods which are based on the functional mock-up interface (FMI) standard (e.g., [62,63]).…”