Limits...
Unsupervised learning of reflexive and action-based affordances to model adaptive navigational behavior.

Weiller D, Läer L, Engel AK, König P - Front Neurorobot (2010)

Bottom Line: The robot learns knowledge of the environment by memorizing the sensory outcome of its motor actions.Differentiating distal and central processing increases the behavioral accuracy of the selected actions and the ability of behavioral adaptation to a changed environment.We propose that the system can canonically be expanded to model other behaviors, using alternative definitions of states and actions.

View Article: PubMed Central - PubMed

Affiliation: Institute of Cognitive Science, Department of Neurobiopsychology, University of Osnabrück Osnabrück, Germany.

ABSTRACT
Here we introduce a cognitive model capable to model a variety of behavioral domains and apply it to a navigational task. We used place cells as sensory representation, such that the cells' place fields divided the environment into discrete states. The robot learns knowledge of the environment by memorizing the sensory outcome of its motor actions. This is composed of a central process, learning the probability of state-to-state transitions by motor actions and a distal processing routine, learning the extent to which these state-to-state transitions are caused by sensory-driven reflex behavior (obstacle avoidance). Navigational decision making integrates central and distal learned environmental knowledge to select an action that leads to a goal state. Differentiating distal and central processing increases the behavioral accuracy of the selected actions and the ability of behavioral adaptation to a changed environment. We propose that the system can canonically be expanded to model other behaviors, using alternative definitions of states and actions. The emphasis of this paper is to test this general cognitive model on a robot in a real-world environment.

No MeSH data available.


Navigational behavior of the robot was investigated by measuring the length of the path to different goals. The direct path, defined as the shortest traversable path from the start point to the goal state (shown as the gray line in the upper part), was used to normalize the length of the robot's path (yellow line) to the goal. The red line corresponds to the length of a path of a simulated robot by taking the topographical distribution of states (geometric transition matrix) into account. The bars represent the median length between different starting and goal states and their standard deviation.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC2871689&req=5

Figure 4: Navigational behavior of the robot was investigated by measuring the length of the path to different goals. The direct path, defined as the shortest traversable path from the start point to the goal state (shown as the gray line in the upper part), was used to normalize the length of the robot's path (yellow line) to the goal. The red line corresponds to the length of a path of a simulated robot by taking the topographical distribution of states (geometric transition matrix) into account. The bars represent the median length between different starting and goal states and their standard deviation.

Mentions: The navigation performance of the robot was evaluated by repeatedly measuring its path to a number of different target sites in the environment. In each of the 20 trials, the robot was placed on one of five possible starting positions and given one of four target locations. In order to directly compare different start-target combinations, we normalized the length of the robot's path by the direct path, which represented the shortest traversable distance from the robot's starting point to the goal state. Figure 4 shows a path traveled by the robot (yellow line) and the corresponding direct path (light gray line). Overall, the robot's median path length across 20 trials was 1.71, with a standard deviation of 0.47. This represents an increase of 71% (±47%) compared to the direct path length. For all configurations of start positions and targets, the robot was able to reach the target in a reasonably short amount of time.


Unsupervised learning of reflexive and action-based affordances to model adaptive navigational behavior.

Weiller D, Läer L, Engel AK, König P - Front Neurorobot (2010)

Navigational behavior of the robot was investigated by measuring the length of the path to different goals. The direct path, defined as the shortest traversable path from the start point to the goal state (shown as the gray line in the upper part), was used to normalize the length of the robot's path (yellow line) to the goal. The red line corresponds to the length of a path of a simulated robot by taking the topographical distribution of states (geometric transition matrix) into account. The bars represent the median length between different starting and goal states and their standard deviation.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC2871689&req=5

Figure 4: Navigational behavior of the robot was investigated by measuring the length of the path to different goals. The direct path, defined as the shortest traversable path from the start point to the goal state (shown as the gray line in the upper part), was used to normalize the length of the robot's path (yellow line) to the goal. The red line corresponds to the length of a path of a simulated robot by taking the topographical distribution of states (geometric transition matrix) into account. The bars represent the median length between different starting and goal states and their standard deviation.
Mentions: The navigation performance of the robot was evaluated by repeatedly measuring its path to a number of different target sites in the environment. In each of the 20 trials, the robot was placed on one of five possible starting positions and given one of four target locations. In order to directly compare different start-target combinations, we normalized the length of the robot's path by the direct path, which represented the shortest traversable distance from the robot's starting point to the goal state. Figure 4 shows a path traveled by the robot (yellow line) and the corresponding direct path (light gray line). Overall, the robot's median path length across 20 trials was 1.71, with a standard deviation of 0.47. This represents an increase of 71% (±47%) compared to the direct path length. For all configurations of start positions and targets, the robot was able to reach the target in a reasonably short amount of time.

Bottom Line: The robot learns knowledge of the environment by memorizing the sensory outcome of its motor actions.Differentiating distal and central processing increases the behavioral accuracy of the selected actions and the ability of behavioral adaptation to a changed environment.We propose that the system can canonically be expanded to model other behaviors, using alternative definitions of states and actions.

View Article: PubMed Central - PubMed

Affiliation: Institute of Cognitive Science, Department of Neurobiopsychology, University of Osnabrück Osnabrück, Germany.

ABSTRACT
Here we introduce a cognitive model capable to model a variety of behavioral domains and apply it to a navigational task. We used place cells as sensory representation, such that the cells' place fields divided the environment into discrete states. The robot learns knowledge of the environment by memorizing the sensory outcome of its motor actions. This is composed of a central process, learning the probability of state-to-state transitions by motor actions and a distal processing routine, learning the extent to which these state-to-state transitions are caused by sensory-driven reflex behavior (obstacle avoidance). Navigational decision making integrates central and distal learned environmental knowledge to select an action that leads to a goal state. Differentiating distal and central processing increases the behavioral accuracy of the selected actions and the ability of behavioral adaptation to a changed environment. We propose that the system can canonically be expanded to model other behaviors, using alternative definitions of states and actions. The emphasis of this paper is to test this general cognitive model on a robot in a real-world environment.

No MeSH data available.