Paper Title
Driverless Vehicles Path Planning and Control System using MPPI and KAN

Abstract
This research introduces few new approaches to improve path planning and control systems for driverless vehicles using Kolmogorov-Arnold Networks (KAN). It focuses on two novel methods: a reinforcement learning based KAN algorithm and a KAN based Model Predictive Path Integral (MPPI) algorithm [12]. Model Predictive Control (MPC) optimises paths by solving constrained optimisation problems, while MPPI improves upon MPC with stochastic sampling for better handling. Usual implementations rely upon multilayer perceptron (MLPs) or mathematical models, which is problematic with complex representations. KAN offers a new approach, forming a structured architecture to learn system equations more effectively than MLPs. This paper also replaces Deep Q-Networks (DQN) with KAN in RL for enhancing and integrating KAN into MPPI to improve real control signals. Simulations show the results of KAN as a transformation tool for autonomous vehicle control and path finding. Keywords - Kolmogorov Arnold Network, Robot Operating System, Multilayer Perceptron, Path Planning, Autonomous Vehicles.