Control Plan Sequence Evaluation Using Learned Future State Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for generating control plans with robot arms using task and motion planning (TAMP) struggle with insufficient constraint conditions, leading to infeasible plans due to the lack of consideration for long-term influences.
Innovation Solution
An information processing apparatus that includes a processor for acquiring state information and plan sequences, generating a state sequence using a learned model, and calculating a success probability of the plan sequence, taking into account long-term influences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If task and motion planning (TAMP) with manually designed constraint conditions is used, then the success rate of the plan can be improved, but it is difficult to cover all necessary constraint conditions leading to infeasible plans
Solution Approach 1:
The patent replaces manual mechanical design of constraint conditions with a neural network-based learned model. The neural network automatically learns appropriate constraint conditions from training data, substituting the manual TAMP constraint design process and enabling the system to handle complex long-term dependencies without requiring explicit manual constraint specification.
Solution Approach 2:
The learned model performs self-service by automatically generating and applying constraint conditions based on input images and state information. Instead of requiring external manual constraint design, the system autonomously determines appropriate constraints through the neural network's learned parameters, making the constraint generation process self-sufficient.
2Ease of operation
If only current image information is considered for control planning, then the processing simplicity is maintained, but long-term influence cannot be considered reducing plan success probability
Solution Approach 1:
The patent applies preliminary action by using the learned model to predict future states and evaluate long-term consequences of control plans before execution. The neural network processes current state information and plan sequences to anticipate future outcomes, enabling the system to consider long-term influences in advance rather than reacting to immediate conditions only.
Solution Approach 2:
The learned model acts as an intermediary between current state information and future outcome evaluation. The neural network receives current images and plan sequences as input, processes them through learned parameters, and outputs predictions about future states and success probabilities, mediating the connection between present conditions and long-term consequences.
Data Source
AI summary
Provided is a technique that makes it possible, as evaluation of a control plan with respect to a control target, to evaluate a plan sequence while taking into consideration long-term influence. An information processing apparatus includes: an acquisition section that acquires state information and a plan sequence, the state information indicating a state of at least one of a control target and an environment, and the plan sequence being a time series of control plans with respect to the control target; a generation section that generates a state sequence using output obtained by inputting the state information and the plan sequence into a learned model, the state sequence being a time series of pieces of state information each indicating a predicted future state; and a calculation section that calculates, using the state sequence, a success probability of the plan sequence which has been acquired by the acquisition section.