Obstacle Probability Map for Autonomous Robots
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current SLAM techniques for obstacle detection in robotics are CPU-intensive and inefficient for detecting mobile obstacles, leading to high complexity and potential CPU availability issues for human interactions in humanoid robots.
Innovation Solution
A method for creating a local map of obstacle probability using a pixel map with predefined limits and dimensions, updated by sensor data from various types of sensors, which maintains low CPU usage and efficiently detects mobile obstacles without additional complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If SLAM techniques are used for obstacle detection and localization, then the robot can locate itself and detect obstacles, but the CPU usage becomes critically high and complexity increases
Solution Approach 1:
The patent segments the environment map into discrete cells with simplified probability values (0, 0.5, or 1) representing obstacle presence, absence, or uncertainty. This cell-based segmentation replaces the continuous, complex probability distributions used in SLAM, significantly reducing computational complexity while maintaining reliable obstacle detection capabilities.
Solution Approach 2:
The patent uses a lightweight, simplified map representation that can be quickly updated and discarded. Instead of maintaining complex SLAM data structures with covariance matrices and landmark positions, the system uses simple probability maps that require minimal computational resources to update and can be rapidly regenerated as needed.
2Loss of information
If SLAM techniques are used to map the environment, then the robot can build a comprehensive map, but the memory footprint increases and CPU availability for human interactions is reduced
Solution Approach 1:
The patent applies local quality by maintaining detailed probability information only in the robot's local vicinity rather than throughout the entire environment. The probability map focuses computational resources on cells near the robot, using simpler representations for distant areas, thereby reducing overall memory footprint and CPU energy consumption while preserving essential environmental information.
Solution Approach 2:
The patent uses partial action by updating only the necessary portions of the probability map based on current sensor observations rather than recomputing the entire map. This selective updating approach maintains environmental information completeness in relevant areas while minimizing CPU energy consumption and memory usage.
3Area of stationary object
If the environment map size increases as the robot discovers new areas, then the coverage improves, but the obstacle detection complexity and CPU usage become unbounded
Solution Approach 1:
The patent implements dynamic map management where the probability map automatically adapts to the robot's current position and orientation. As the robot moves to new areas, the map dynamically updates to cover relevant regions while maintaining a fixed resolution and cell size. This dynamic approach ensures map coverage grows with exploration while keeping obstacle detection complexity bounded through the simplified probability representation.
Data Source
AI summary
A method to compute a pixel map of probabilities of absence and presence of an obstacle in the environment of an autonomous robot comprising at least one sensor grouped in at least one set of sensors that detect obstacles of similar types and have its own sensor initial map of probability of absence of obstacles is provided. The method comprises the steps of initializing a map around the robot and attached to the robot with a predefined value of probability of absence or presence of an obstacle, acquiring data representative of the absence or presence of an obstacle around the robot from at least one sensing procedure, and concurrently updating values of probabilities using data from sensing procedure and modifying the probabilities of absence or presence of obstacle from previous observations to a value closer to a predefined value.