<p>A multi-strategy improved sand cat swarm optimization algorithm (CAM-SCSO) is proposed for time-optimal trajectory planning of a fruit-picking robotic arm operating in unstructured orchard environments, addressing poor time efficiency and the tendency of conventional methods to become trapped in local optima. To overcome the limitations of the original SCSO in such scenarios, a chaos–opposition strategy is adopted for population initialization to enhance diversity, an adaptive differential perturbation mechanism is incorporated into the search phase to improve exploration capability and accelerate convergence, and a multi-angle hybrid localization strategy is applied in the predation phase to boost convergence accuracy. Compared with seven algorithms on the CEC2022 test suite, CAM-SCSO is shown to achieve the best performance on 90% of the benchmark functions, demonstrating superior accuracy, speed, and stability. Using the Xarm6 six-axis robotic arm as the platform, its kinematic model was constructed via Denavit–Hartenberg (D–H) parameters, and trajectories were interpolated by 3–5–3 polynomial interpolation; when applied to time-optimal trajectory planning, a 23.59% reduction in motion time is achieved, along with smooth and continuous kinematic curves. Finally, the effectiveness of the improved algorithm is further validated by simulations in MATLAB/ROS and by physical tests in a fruit tree environment.</p>