Aerospace Instrument-Making Annotation << Back Joint Trajectory Planning of Dynamic Objects based on Reinforcement Learning Zhu Han, Peng Sikun, Selezneva M.S. A visual reinforcement learning environment based on the Soft Actor-Critic (SAC) algorithm is developed for efficient joint trajectory planning of unmanned aerial vehicles (UAVs) and unmanned cars. Experimental results confirm that the trained system successfully plans a route to a target location, effectively avoiding obstacles along the way. The SAC algorithm demonstrates high flexibility by adapting to a different number of UAVs and outperforms the random strategy and MADDPG (Multi-Agent Deep Deterministic Policy Gradient) algorithm, which provides synergy in the agents' actions. Keywords: UAV, unmanned vehicle, reinforcement learning, trajectory planning, heterogeneous system, Markov process. DOI: 10.25791/aviakosmos.8.2025.1501 Pp. 56-63.