SMARTERS - Simulation of Mowing Agents in a Tile-based Environments with Realistic Space

Description

SMARTERS is a customizable simulator for testing autonomous robot performance in complex environments. Built on the Mesa framework, it combines a continuous space with a discrete tile grid, allowing for both fine-grained motion tracking and statistical analysis of robot behavior. The simulator supports blocked zones, isolated areas, different movement and cutting models, and JSON-based environment configuration.

Environment Representation

Dual-Space Model

SMARTERS uses two parallel representations of space:

• Continuous Space (ContinuousSpace): The robot moves using real coordinates, allowing smooth transitions and directional flexibility. This is the space where motion logic is computed.
• Discrete Grid (MultiGrid): A tiled representation of the environment used for tracking visits and displaying features (e.g., grass, lines, obstacles). Each tile logs how many times the robot has passed over it.

This dual-layer model allows for accurate simulation of robot behavior and robust analysis of area coverage.

Tile Grid

Tiles represent discrete areas of the environment, each capable of holding multiple agents or resources:

• Agents: Dynamic entities (e.g., robots), implemented using Mesa’s Agent class.
• Resources: Static objects such as grass, guiding lines, isolated areas, blocked areas, and entry points.

Key Resources

• Grass Tiles
• Guiding Lines
• Isolated Areas
• Blocked Areas (Square or Circular)
• Openings (for controlled access to isolated areas)

Obstacle and Area Types

Blocked Areas

Inaccessible zones for the robot. Defined as:

• SquaredBlockedArea
• CircledBlockedArea

They represent real-world obstructions like buildings or pools and can be placed manually or generated randomly.

Isolated Areas

Zones that the robot can enter only through predefined Opening tiles. Managed with the IsolatedArea class. Their size, shape, and placement can be manually configured or randomized.

Robot Behavior Models

Bounce Model

Determines how the robot reacts upon hitting an obstacle:

• Ping Pong: Reflects the direction (bounces back).
• Random: Attempts upper-left movement first, falling back on other directions if blocked.

Cutting Model

Represents the robot's grass-cutting logic:

• The robot "cuts" by traversing tiles and incrementing a counter.
• In random cutting mode, it selects a random direction and proceeds until blocked, potentially crossing multiple tiles diagonally.

Configuration

JSON Input

SMARTERS uses a configuration file in JSON format. Two formats are supported:

• Structure 1: Full configuration with:

  • Robot specs (type, speed, cutting mode, etc.)
  • Environment settings (dimensions, blocked area counts, etc.)
  • Simulation parameters (tile size, number of cycles, etc.)

• Structure 2: Grid definition using Cartesian coordinates, manually specifying:

  • Blocked areas
  • Isolated areas
  • Openings

Example Command

python main.py --d data_file.json

Ensure the View/ folder is writable and the following libraries are installed: mesa, pandas, numpy, matplotlib, seaborn.

Plugin Support

SMARTERS supports modular extensions through Python-based plugins.

Runtime Options

• --e: Environment plugin (e.g., environment_plugin.py)
• --r: Robot plugin (e.g., robot_plugin.py)
• --d: JSON configuration file

Example:

python main.py --e environment_plugin.py --r robot_plugin.py --d data_file.json

Simulation Cycle

Each simulation follows these steps:

1. Map generation and base station placement
2. Placement of blocked and isolated areas
3. Continuous robot motion across the space
4. Tile-based logging of robot presence
5. Repeat across multiple cycles until completion

The robot has limited autonomy per cycle. Once exhausted, it recharges before resuming activity.

Output

CSV Exports

• Grid Representation: Shows tile types across the grid.
• Cycle Files (cycle_i.csv): Record how many times the robot passed over each tile in cycle i.

Visual Analysis

• Histograms: Distribution of tile traversals (X = number of crossings, Y = number of tiles).
• Heatmaps: Color-coded views of high and low traversal zones.

Note: Warnings in the terminal about agents occupying the same tile can be ignored.

Summary

SMARTERS offers a flexible, powerful framework for simulating autonomous mowing robots in rich, user-defined environments. With support for continuous motion, discrete logging, customizable behaviors, and plugin-based extension, it enables detailed performance evaluation for research or development purposes.