Efficient Goal Estimation for Intelligent Systems

TECHNOLOGY DESCRIPTION

To build intelligent systems and/or robots or understand the behaviour of decision makers, it is important to estimate the goal of human or animal subject behaviour by observation.
This technology is a modelfree algorithm to estimate the goal of decision makers from their behavioural data with reduced memory usage and computational cost. In addition, action selection policies can be recovered more effectively than other known methods as shown in the figure.

BENEFITS

- Small datasets for analysis
- Reduced memory usage
- Reduced computational cost

LIMITATIONS

As is the case with the prior-art, a certain minimum number of observations is required to recover action selection policies.

APPLICATIONS

- Interpretation of human behavior
- Analysis of web experience
- Robot control by imitation

MATERIALS