Key Engineering Takeaways
- •The SLAM dilemma: To build a map, the robot must know where it is; to localize itself, the robot needs an accurate map.
- •Frontend computes local incremental scan-to-scan constraints; Backend optimizes the entire trajectory graph to minimize global residual error.
- •Loop closure detects when the robot revisits a previously mapped location, eliminating accumulated dead-reckoning drift instantaneously.
Prerequisites
- • Linear algebra
- • Probability & non-linear least squares
The Chicken-and-Egg SLAM Dilemma
When an autonomous mobile robot enters an unknown warehouse or building, it starts with an empty map and unknown position. As the robot moves, sensor noise causes odometry errors to compound quadratically over distance.
**Simultaneous Localization and Mapping (SLAM)** solves this by maintaining a joint probability distribution over both the robot trajectory and the landmark locations:
$$P(x_{1:t}, m \mid z_{1:t}, u_{1:t})$$
Backend: Non-Linear Pose Graph Optimization
In modern Graph-SLAM, robot poses form the **nodes** of a graph, and sensor observations (odometry and scan matching) form the **edges** (spatial constraints).
The backend minimizes the sum of squared Mahalanobis distance residuals over all constraints:
$$\mathbf{x}^* = \arg\min_{\mathbf{x}} \sum_{i,j} \mathbf{e}_{ij}(\mathbf{x}_i, \mathbf{x}_j)^T \boldsymbol{\Omega}_{ij} \mathbf{e}_{ij}(\mathbf{x}_i, \mathbf{x}_j)$$
Using solvers like **Ceres Solver** or **g2o**, non-linear optimization with Levenberg-Marquardt rapidly converges to the globally consistent map.
Tags:#SLAM#Graph-SLAM#Scan Matching#Loop Closure#Ceres Solver#g2o#Pose Graph