Accurately estimating the position and speed of pedestrians is crucial for autonomous vehicles to operate safely in congested environments. LiDAR technology is highly accurate in acquiring the distance to objects, but pedestrian point clouds may be occluded in practical situations. To address this issue, we propose a method that represents pedestrian existence as a probability distribution using kernel density estimation (KDE), accommodating occlusion with a Particle Filter (PF).

This paper presents a robust pedestrian trajectory estimation and combination leveraged by the Hungarian method in crowded environments. Traditional Bayesian filters, such as the Joint Probabilistic Data Association Filter (JPDAF), sometimes fail to track pedestrians continuously due to persistent occlusion and abrupt movements for collision avoidance. In this study, we have improved the success rate of JPDAF-based pedestrian tracking by combining broken trajectories using the Hungarian method. To validate our approach, we conducted multiple-pedestrian tracking experiments in a crowded environment observed by two LiDARs. We confirmed that, even when JPDAF lost track of the target due to frequent occlusion, the proposed method reflecting the direction of the pedestrian flow improved the tracking success rate substantially by reconnecting isolated trajectories.

Leg/wheel mobile robots are expected to play an active role in environments with many obstacles because their leg and wheel mechanisms allow them to adapt to uneven terrain and move efficiently. In this study, for a planar leg/wheel mobile robot, we develop an obstacle avoidance control that combines Model Predictive Control (MPC) based on Markov Chain Monte Carlo (MCMC), a sample-based solution method, and MPC based on a numerical solution to the Euler-Lagrange equations. Specifically, the optimal input is calculated utilizing both MCMC samples and the C/GMRES, and then the samples for the next control cycle are generated through resampling. This approach generates a sub-optimal control input sequence while searching for a global optimal solution, which anticipates that the robot prevents from stacking into a local optimum. The effectiveness of the proposed method is confirmed by comparing it with the MCMPC or C/GMRES methods, respectively, and implementing it into the onboard computer equipped with the actual robot.

This paper proposes an assistant and training system for controlling an unmanned helicopter. The unmanned helicopter does not have any sensors which measure its position or posture. Stationary cameras are placed on the flight field. The helicopter is controlled by using a visual servo technique as follows. An operator steers the helicopter using sticks on a hand-held input device. The sticks make reference signals. The assistant system designs control signals such that the helicopter tracks the reference signals. The proposed system has the following four functions: automatic takeoff and landing, control channel selection, flight in a desired area, and automatic motion generation. They enables beginners to control an unmanned helicopter. The system provides real actions of unmanned helicopters. This is the main difference from flight simulators.