Autonomous driving controllers with neuromorphic spiking neural networks

Abstract

Autonomous driving is one of the hallmarks of artificial intelligence. Neuromorphic (brain-inspired) control is posed to significantly contribute to autonomous behavior by leveraging spiking neural networks-based energy-efficient computational frameworks. In this work, we have explored neuromorphic implementations of four prominent controllers for autonomous driving: pure-pursuit, Stanley, PID, and MPC, using a physics-aware simulation framework. We extensively evaluated these models with various intrinsic parameters and compared their performance with conventional CPU-based implementations. While being neural approximations, we show that neuromorphic models can perform competitively with their conventional counterparts. We provide guidelines for building neuromorphic architectures for control and describe the importance of their underlying tuning parameters and neuronal resources. Our results show that most models would converge to their optimal performances with merely 100–1,000 neurons. They also highlight the importance of hybrid conventional and neuromorphic designs, as was suggested here with the MPC controller. This study also highlights the limitations of neuromorphic implementations, particularly at higher (> 15 m/s) speeds where they tend to degrade faster than in conventional designs.

Continuous adaptive nonlinear model predictive control using spiking neural networks and real-time learning

“…Adaptive spiking neurons were used to adaptively change the hexapod's body leveling during transversal over multi-leveled terrain. In the context of autonomous driving, SNNs were recently utilized to mirror traditional autonomous driving controllers such as the pure-pursuit, Stanley, PID, and MPC [14]. This allows them to be incorporated into continuous time control, offering a low-power alternative.…”

An organic brain-inspired platform with neurotransmitter closed-loop control, actuation and reinforcement learning

“…In addition, neuromorphic architectures, by exploiting neural primitives, may enable real-time interaction with the surrounding environment. Indeed, state-of-the-art neuromorphic controllers may either leverage on neuromorphic models to determine a control law, 26,27 or they can rely on spiking neural networks (SNNs) to implement long-standing control laws, such as proportional, integral and derivative (PID) controllers on silicon neuromorphic chips. 28,29 Crucially, such approaches still fail in recapitulating the autonomous adaptation that characterizes neural processing.…”

Intelligent Robotics in Pediatric Cooperative Neurorehabilitation: A Review

“…Recent advances in the utilization of neuromorphic designs to provide adaptive robotic control show great promise in various applications such as classical inverse kinematic calculations in joint-based systems featuring low [123] and high degrees of freedom [170], as well as in free-moving autonomous vehicular systems [171]. It was recently implemented for the first time in a clinical rehabilitation framework where a neuromorphically controlled framework was used to control a robotic arm mounted on a wheelchair, providing accurate responsive control with low energy requirements and a high level of adaptability [137].…”