The following sections summairze the Grid-based FastSLAM algorithm which is one instance of FastSLAM. This algorithm estimates the trajectory of a mobile robot while simultaneously creating a grid map of the environment. Grid-based FastSLAM is combination of a particle filter such as Adaptive Monte Carlo Localization (amcl) and a mapping algorithm such as occupancy grid mapping.



Recursive State Estimation

Robot Environment Interaction


Environments are characterized by state. State that change over time is called dynamic state, e.g., moving people or other vehicles. Static state is non-changing state, such as the location of walls in (most) buildings. The state also includes variables regarding the robot itself, such as its pose, velocity, whether or not its sensors are functioning correctly and so on. State is denoted by $x$ and the state at time $t$ by $x_t$. Typical state variables include:

  • Robot pose
  • In robot manipulation, the pose includes variables for the configuration of robot’s actuators e.g., joint angles. Degrees of freedom is related to the kinematic state of a robot.
  • Robot velocity and velocities of its joints are commonly referred to as dynamic state.
  • Locations and features of surrounding objects in the environment are also state variables e.g., trees, walls. In some problems, objects will assume the form of landmarks, which are distinct, staationary features of the environment that can be recognized reliably.
  • Location and velocity of moving objects and people are also potential state variables.
  • Broken sensors or level of battery can be state variables.
Environment Interaction
Probabilistic Generative Laws
Belief Distributions

Bayes Filter

  • Bayes Filter Algorithm
  • Markov Assumption

Gaussian Filters

Nonparametric Filters

Robot Motion

Robot Perception


Mobile Robot Localization: Markov and Gaussian

Mobile Robot Localization: Grid And Monte Carlo


Occupancy Grid Mapping

Simultaneous Localization and Mapping

The GraphSLAM Algorithm

The Sparse Extended Information Filter

The FastSLAM Algorithm

Planning and Control

Markov Decision Processes

Partially Observable Markov Decision Processes

Approximate POMDP Techniques