Biased-MPPI improves Model Predictive Path Integral (MPPI) control by allowing arbitrary sampling distributions, enhancing efficiency, robustness, and convergence while reducing local minima. The method integrates classical and learning-based ancillary controllers for more informative sampling and control fusion. Simulated and real-world experiments show that Biased-MPPI outperforms traditional MPPI in handling dynamic environments, model uncertainties, and unexpected events. It achieves better performance by biasing the sampling distribution towards controllers that improve robustness and efficiency. In a rotary inverted pendulum experiment, Biased-MPPI successfully swings up the pendulum and avoids collisions. In multi-vehicle scenarios, Biased-MPPI reduces collisions and improves navigation in constrained environments. The method also shows improved performance in real-world motion planning, avoiding collisions when unexpected obstacles appear. The approach introduces a potential bias towards slower trajectories due to certain control strategies, but overall improves safety, performance, and sample efficiency. The paper concludes that Biased-MPPI is a promising solution for complex multi-modal problems, particularly in autonomous systems.Biased-MPPI improves Model Predictive Path Integral (MPPI) control by allowing arbitrary sampling distributions, enhancing efficiency, robustness, and convergence while reducing local minima. The method integrates classical and learning-based ancillary controllers for more informative sampling and control fusion. Simulated and real-world experiments show that Biased-MPPI outperforms traditional MPPI in handling dynamic environments, model uncertainties, and unexpected events. It achieves better performance by biasing the sampling distribution towards controllers that improve robustness and efficiency. In a rotary inverted pendulum experiment, Biased-MPPI successfully swings up the pendulum and avoids collisions. In multi-vehicle scenarios, Biased-MPPI reduces collisions and improves navigation in constrained environments. The method also shows improved performance in real-world motion planning, avoiding collisions when unexpected obstacles appear. The approach introduces a potential bias towards slower trajectories due to certain control strategies, but overall improves safety, performance, and sample efficiency. The paper concludes that Biased-MPPI is a promising solution for complex multi-modal problems, particularly in autonomous systems.