Benchmarking Classical Path Planning Algorithms for Excavator End-Effector Motion in Physics-Based Simulation

Abstract Reliable motion planning is a key requirement for the autonomous operation of construction machinery, particularly for excavator manipulators operating in constrained workspaces. This paper presents a simulation-based benchmark of four classical path planning algorithms – A*, Rapidly-exploring Random Trees (RRT), RRT*, and Probabilistic Roadmaps (PRM) – for excavator end-effector motion in a container-entry task representative of pick-and-place manipulation. Experiments are conducted in the NVIDIA IsaacSim physics-based simulation environment using a CAD-derived rigid-body excavator model. Both full three-dimensional and planar start–goal formulations are evaluated together with parametric sweep conditions across 940 successful automated trials. The evaluation records planning time, trajectory geometry and tracking accuracy. Actuator effort metrics, such as torque and cumulative mechanical work estimates are also provided and interpreted as relative indicators of trajectory geometric demand. Jacobian condition numbers and Yoshikawa manipulability indices are monitored to assess inverse kinematics (IK) numerical stability. The results demonstrate clear performance differentiation between planners. A* planning time is reduced from 26.0 s in full 3D to 1.1 s under the planar formulation with negligible change in path length. Conventional RRT produces trajectories over 60% longer than A* and imposes cumulative actuator demand exceeding three times that of RRT* or PRM, while planar RRT* and planar PRM achieve the lowest combined effort with compact trajectories. IK diagnostics vary by less than 5% across planners, confirming that performance differences reflect trajectory geometry rather than kinematic conditioning. The study provides practical algorithm selection guidance and a reproducible simulation foundation for future sim-to-real transfer on construction machinery.