Conference Paper

Multilayered reinforcement learning for complicated collision avoidance problems

RIKEN, Inst. of Phys. & Chem. Res., Saitama
DOI: 10.1109/ROBOT.1998.680648 Conference: Robotics and Automation, 1998. Proceedings. 1998 IEEE International Conference on, Volume: 3
Source: IEEE Xplore

ABSTRACT

We have proposed the collision avoidance methods in a multirobot
system based on the information exchanged by the “LOCISS: Locally
Communicable Infrared Sensory System”, which is developed by the
authors. One of the problems in the LOCISS based methods is that the
number of situations which should be considered increases very much when
the number of the robots and stationary obstacles in the working
environment increases. In order to reduce the required computational
power and memory capacity for such a large number of situations, we
propose, in this paper, a multilayered reinforcement learning scheme to
acquire appropriate collision avoidance behaviors. The feasibility and
the performance of the proposed scheme is examined through the
experiment using actual mobile robots

0 Followers
 · 
3 Reads
  • Source
    • "The fuzzy system is automatically designed to train the neural network weights. Fujii et al. [9] suggested a multi-layered methodology for collision-free navigation via reinforcement learning. However, the planned vehicle motions using learning based approaches are not optimal, particularly at the initial learning stage. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Real-time safety aware navigation of an intelligent vehicle is one of the major challenges in intelligent vehicle systems. Many studies have been focused on the obstacle avoidance to prevent an intelligent vehicle from approaching obstacles “too close” or “too far”, but difficult to obtain an optimal trajectory. In this paper, a novel biologically inspired neural network methodology with safety consideration to realtime collision-free navigation of an intelligent vehicle with safety consideration in a non-stationary environment is proposed. The real-time vehicle trajectory is planned through the varying neural activity landscape, which represents the dynamic environment, in conjunction of a safety aware navigation algorithm. The proposed model for intelligent vehicle trajectory planning with safety consideration is capable of planning a real-time “comfortable” trajectory by overcoming the either “too close” or “too far” shortcoming. Simulation results are presented to demonstrate the effectiveness and efficiency of the proposed methodology that performs safer collision-free navigation of an intelligent vehicle.
    Full-text · Conference Paper · Jun 2014
  • Source
    • "Svestka [14] proposed a probabilistic learning approach to motion planning of a mobile robot, which involves a learning phase and a query phase and uses a local method to compute feasible paths for the robot. Fujii [15] proposed a multilayer reinforcement learning model for path planning of multiple mobile robots. However, the planned robot motion using learning based approaches is not efficient and is computationally expensive, especially in its initial learning phase. "
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper a biologically inspired neural dynamics and map planning based approach are simultaneously proposed for AUV (Autonomous Underwater Vehicle) path planning and obstacle avoidance in an unknown dynamic environment. Firstly the readings of an ultrasonic sensor are fused into the map using the D-S (Dempster-Shafer) inference rule and a two-dimensional occupancy grid map is built. Secondly the dynamics of each neuron in the topologically organized neural network is characterized by a shunting equation. The AUV path is autonomously generated from the dynamic activity landscape of the neural network and previous AUV location. Finally, simulation results show high quality path optimization and obstacle avoidance behaviour for the AUV.
    Full-text · Article · Mar 2014 · International Journal of Advanced Robotic Systems
  • Source
    • "Multi-robot system (MRS) has received more and more attention because of its broad application prospect, which has several research platforms including formation [1], foraging [2], prey-pursuing [3] [4], and robot soccer [5] [6] [7]. Robot soccer is associated with robot architecture, cooperation, decision making, planning, modeling, learning, vision tracking algorithm , sensing, and communication, which owns all the key features of MRS. "
    [Show abstract] [Hide abstract]
    ABSTRACT: This paper proposes a novel multiagent reinforcement learning (MARL) algorithm Nash- learning with regret matching, in which regret matching is used to speed up the well-known MARL algorithm Nash- learning. It is critical that choosing a suitable strategy for action selection to harmonize the relation between exploration and exploitation to enhance the ability of online learning for Nash- learning. In Markov Game the joint action of agents adopting regret matching algorithm can converge to a group of points of no-regret that can be viewed as coarse correlated equilibrium which includes Nash equilibrium in essence. It is can be inferred that regret matching can guide exploration of the state-action space so that the rate of convergence of Nash- learning algorithm can be increased. Simulation results on robot soccer validate that compared to original Nash- learning algorithm, the use of regret matching during the learning phase of Nash- learning has excellent ability of online learning and results in significant performance in terms of scores, average reward and policy convergence.
    Full-text · Article · Aug 2013 · Mathematical Problems in Engineering
Show more