Univector Field Research Articles

Univector field method-based multi-agent navigation for pursuit problem in obstacle environments

The pursuit problem is a well-known problem in computer science. In this problem, a group of predator agents attempt to capture a prey agent in an environment with various obstacle types, partial observation, and an infinite grid-world. Predator agents are applied algorithms that use the univector field method to reach the prey agent, strategies for avoiding obstacles and strategies for cooperation between predator agents. Obstacle avoidance strategies are generalized and presented through strategies called hitting and following boundary (HFB); trapped and following shortest path (TFSP); and predicted and following shortest path (PFSP). In terms of cooperation, cooperation strategies are employed to more quickly reach and capture the prey agent. Experimental results are shown to illustrate the efficiency of the method in the pursuit problem.

Real-time Footstep Planning and Following for Navigation of Humanoid Robots

This paper proposes novel real-time footstep planning and following methods for the navigation of humanoid robots. A footstep command is defined by a walking direction and step lengths for footstep planning. The walking direction is determined by a uni-vector field navigation method, and the allowable yawing range caused by hardware limitation is considered. The lateral step length is determined to avoid collisions between the two legs while walking. The sagittal step length is modified by a binary search algorithm when collision occurs between the robot body and obstacles in a narrow space. If the robot body still collides with obstacles despite the modification of the sagittal step length, the lateral step length is shifted at the next footstep. For footstep following, a walking pattern generator based on a 3-D linear inverted pendulum model is utilized, which can generate modifiable walking patterns using the zero-moment point variation scheme. Therefore, it enables a humanoid robot to follow the footstep command planned for each footstep. The effectiveness of the proposed method is verified through simulation and experiment.

Dyna-QUF: Dyna-Q based univector field navigation for autonomous mobile robots in unknown environments

A novel approach was presented to solve the navigation problem of autonomous mobile robots in unknown environments with dense obstacles based on a univector field method. In an obstacle-free environment, a robot is ensured to reach the goal position with the desired posture by following the univector field. Contrariwise, the univector field cannot guarantee that the robot will avoid obstacles in environments. In order to create an intelligent mobile robot being able to perform the obstacle avoidance task while following the univector field, Dyna-Q algorithm is developed to train the robot in learning moving directions to attain a collision-free path for its navigation. Simulations on the computer as well as experiments on the real world prove that the proposed algorithm is efficient for training the robot in reaching the goal position with the desired final orientation.

Univector Field Method based Multi-Agent Navigation for Pursuit Problem

This paper presents a new approach to solve the pursuit problem based on a univector field method. In our proposed method, a set of eight agents works together instantaneously to find suitable moving directions and follow the univector field to pursue and capture a prey agent by surrounding it from eight directions in an infinite grid-world. In addition, a set of strategies is proposed to make the pursuit problem more realistic in the real world environment. This is a general approach, and it can be extended for an environment that contains static or moving obstacles. Experimental results show that our proposed algorithm is effective for the pursuit problem.

AN EVOLUTIONARY OPTIMIZED FOOTSTEP PLANNER FOR THE NAVIGATION OF HUMANOID ROBOTS

In this paper, an evolutionary optimized footstep planner for the navigation of humanoid robots is proposed. A footstep planner based on a univector field navigation method is proposed to generate a command state (CS) as an input to a modifiable walking pattern generator (MWPG) at each footstep. The MWPG generates associated trajectories of every leg joint to follow the given CS. In order to satisfy various objectives in the navigation, the univector fields are optimized by evolutionary programming. The three objectives, shortest elapsed time to get to a destination, safety without obstacle collision, and less energy consumption, are considered with mechanical constraints of a real humanoid robot, that is, the maximum step length and allowable yawing range of the feet. The effectiveness of the proposed algorithm is demonstrated through both computer simulation and experiment for a small-sized humanoid robot, HanSaRam-IX.

Evolutionary Multiobjective Footstep Planning for Humanoid Robots

This paper proposes a novel evolutionary multiobjective footstep planner for humanoid robots. First, a footstep planner using a univector field navigation method is proposed to provide a command state (CS), which is to be an input of a modifiable walking pattern generator (MWPG) at each footstep. Then, the MWPG generates corresponding trajectories for every leg joint of the humanoid robot at each footstep to follow the CS. Second, a multiobjective evolutionary algorithm (MOEA) is employed to optimize the univector fields satisfying multiple objectives in navigation. Finally, a preference-based selection algorithm based on a fuzzy measure and fuzzy integral is proposed to select the preferred one out of various nondominated solutions obtained by the MOEA. The effectiveness of the proposed evolutionary multiobjective footstep planner is demonstrated through computer simulations for a simulation model of a small-sized humanoid robot, HanSaRam-VIII.

Evolutionary programming-based univector field navigation method for past mobile robots

Most of navigation techniques with obstacle avoidance do not consider the robot orientation at the target position. These techniques deal with the robot position only and are independent of its orientation and velocity. To solve these problems this paper proposes a novel univector field method for fast mobile robot navigation which introduces a normalized two dimensional vector field. The method provides fast moving robots with the desired posture at the target position and obstacle avoidance. To obtain the sub-optimal vector field, a function approximator is used and trained by evolutionary programming. Two kinds of vector fields are trained, one for the final posture acquisition and the other for obstacle avoidance. Computer simulations and real experiments are carried out for a fast moving mobile robot to demonstrate the effectiveness of the proposed scheme.

Modular Q-learning based multi-agent cooperation for robot soccer

In a multi-agent system, action selection is important for the cooperation and coordination among agents. As the environment is dynamic and complex, modular Q-learning, which is one of the reinforcement learning schemes, is employed in assigning a proper action to an agent in the multi-agent system. The architecture of modular Q-learning consists of learning modules and a mediator module. The mediator module of the modular Q-learning system selects a proper action for the agent based on the Q-value obtained from each learning module. To obtain better performance, along with the Q-value, the mediator module also considers the state information in the action selection process. A uni-vector field is used for robot navigation. In the robot soccer environment, the effectiveness and applicability of modular Q-learning and the uni-vector field method are verified by real experiments using five micro-robots.