Performance Comparison


Results collected from runs on a physical Turtlebot4 across simple and complex maze environments with static, added, and removed obstacles.

Environment Obstacle Type Algorithm Time (s) Distance (m)
Simple Environment
SimpleStatic A* 40.434.04
SimpleStatic D* Lite 40.14.19
SimpleStatic Jump A* 39.644.10
SimpleStatic DWA (w costmap) 14.983.88
SimpleStatic Hybrid 15.343.93
SimpleAdd A* FAILFAIL
SimpleAdd D* Lite 49.244.23
SimpleAdd Jump A* FAILFAIL
SimpleAdd DWA (w costmap) 15.793.83
SimpleAdd Hybrid 15.343.86
SimpleRemove A* 40.914.14
SimpleRemove D* Lite 50.224.11
SimpleRemove Jump A* 39.014.19
SimpleRemove DWA (w costmap) 14.813.73
SimpleRemove Hybrid 15.123.88
Complex Environment
ComplexStatic A* 44.894.75
ComplexStatic D* Lite 64.84.88
ComplexStatic Jump A* 47.825.17
ComplexStatic DWA (w costmap) FAILFAIL
ComplexStatic Hybrid 18.94.59
ComplexAdd A* 44.6314.261
ComplexAdd D* Lite 73.25.51
ComplexAdd Jump A* FAILFAIL
ComplexAdd DWA (w costmap) FAILFAIL
ComplexAdd Hybrid FAILFAIL
ComplexRemove A* 41.384.67
ComplexRemove D* Lite 59.965.02
ComplexRemove Jump A* 51.155.19
ComplexRemove DWA (w costmap) 14.543.42
ComplexRemove Hybrid 18.564.54

1 A* Complex/Add partially succeeded on one run. It made it to the goal through brute force of pushing walls.

Pros & Cons


Qualitative trade-offs observed across all five algorithms.

Global Planner

A* (A-Star)

Pros
  • Guaranteed optimal path
  • Predictable, deterministic behavior
  • Well-understood and well-documented
  • Can stack with local planners (DWA)
Cons
  • No dynamic replanning — must restart on new obstacles
  • Can be slow on very large maps
  • Requires fully-known map before planning
Dynamic Replanner

D* Lite

Pros
  • Efficient incremental replanning
  • Handles dynamic and unknown environments
  • Only recomputes affected path segments
  • Strong performance in partially-known maps
Cons
  • More complex to implement and tune
  • Higher overhead than A* in static environments
  • Can produce non-smooth paths
Global Planner

Jump Point Search (JPS)

Pros
  • Much faster than A* on open uniform grids
  • Same path optimality as A*
  • Dramatic reduction in node expansions
Cons
  • Only works correctly on uniform-cost grids
  • No dynamic replanning support
  • Benefits diminish on densely-occupied maps
Reactive Controller

DWA (Dynamic Window Approach)

Pros
  • Reactive real-time obstacle avoidance
  • Respects robot kinematic constraints
  • Configurable cost function for tuning behavior
  • No prior map required (reactive only)
Cons
  • Can get stuck in local minima
  • No global path optimality guarantees
  • Sensitive to weight tuning
  • May struggle in narrow corridors
Hybrid Stack

Hybrid A* + DWA

Pros
  • Global optimality + local reactivity
  • Handles dynamic obstacles along planned path
  • Mirrors production Nav2 architecture
  • More robust than either alone
Cons
  • Higher system complexity and tuning overhead
  • Two nodes must coordinate correctly
  • DWA may deviate from A* plan in cluttered spaces