Semantic Map Motion Planning for Mobile Robot Path Generalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for predicting mobile robot trajectories in dynamic environments struggle to generalize to new environments and heterogeneous datasets, often failing to converge to optimal reward function parameters due to differences in empirical feature counts across various semantic maps.
Innovation Solution
A computer-implemented method that uses semantic maps to determine reward function parameters by computing differences between expected and empirical mean feature counts, updating these parameters iteratively, and employing a backward-forward algorithm to simulate trajectories and infer occupancy maps, allowing adaptation to diverse environments without additional adaptation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional methods use empirical feature counts from training data to determine reward function parameters, then the method works for the training environment, but it fails to generalize to new environments with different geometries and heterogeneous datasets
Solution Approach 1:
The patent transforms the reward function parameter determination from a data-dependent process to a geometry-independent process by changing the fundamental parameters used: instead of using empirical feature counts from training trajectories, the method uses only semantic map geometries and topological relationships. This parameter transformation enables generalization to new environments while maintaining reliable convergence through mathematically grounded geometric computations.
Solution Approach 2:
The patent extracts and removes the dependency on training data empirical feature counts from the reward function determination process. By taking out this environment-specific empirical information and replacing it with universal geometric and topological features of semantic maps, the method achieves environment-independent parameter determination that generalizes across heterogeneous datasets.
2Measurement precision
If the reward function is learned from demonstrated behaviors in specific environments, then optimal motion policy is achieved for those environments, but the method does not generalize to new environments or heterogeneous datasets
Solution Approach 1:
The patent creates a universal reward function parameter determination method that works across multiple environments and heterogeneous datasets. By using geometric and topological features that are invariant to specific environment characteristics, the same methodology can be applied universally to determine optimal motion policies in diverse settings without retraining or environment-specific adaptation.
Solution Approach 2:
The patent changes the fundamental parameters from environment-specific empirical features to universal geometric and topological descriptors. This parameter transformation allows the reward function to maintain high measurement precision for optimal motion policy while achieving broad adaptability across different environments through geometry-based invariance.
3Reliability
If semantic maps are used to encode reward function features, then future human motion prediction is improved, but the method oscillates between local minima and fails to converge
Solution Approach 1:
The patent introduces a feedback mechanism where the reward function parameters are determined through iterative optimization using geometric and topological features of semantic maps. The feedback loop continuously refines parameters by comparing predicted trajectories with actual human motion patterns while maintaining stability through geometry-based constraints that prevent oscillation between local minima.
Solution Approach 2:
The patent changes the parameter space from continuous empirical feature counts to discrete geometric and topological descriptors. This parameter transformation stabilizes the optimization process by reducing sensitivity to initialization and preventing oscillation, while maintaining reliable human motion prediction accuracy through the expressive power of geometric features.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a computer-implemented method for determining a motion trajectory for a mobile robot based on an occupancy prior indicating probabilities of presence of dynamic objects and/or individuals (2) in a map of an environment (E), wherein the occupancy priors are determined by a reward function defined by reward function parameters (θ); the determining of the reward function parameters (θ) comprising the steps of: - providing (S1) a number (Bm) of semantic maps; - providing (S1) a number (Bt) of training trajectories for each of the number (Bm) of semantic maps; - computing (S6-S13) -a gradient as a difference between an expected mean feature count (f̂θ) and an empirical mean feature count (f) depending on each of the number of semantic maps and on each of the number of training trajectories, wherein the empirical mean feature count (f) is the average number of features accumulated over the provided training trajectories of the semantic maps, wherein the expected mean feature count (f̂θ) is the average number of features accumulated by trajectories generated depending on the current reward function parameters (θ); - updating (S14) the reward function parameters (θ) depending on the gradient.