Autonomous navigation in complex outdoor terrains presents critical challenges for unmanned ground vehicles (UGVs) due to the inherent disconnect between global mapping and real-time sensor feedback. This work proposes a hybrid framework that integrates low-resolution Digital Elevation Model (DEM) data with real-time LiDAR-based obstacle detection and terrain analysis for efficient path planning. A global path is initially computed using a preprocessed DEM-based A* algorithm. Subsequently, local sensor data drives adaptive path correction, enabling the UGV to negotiate sudden environmental changes while maintaining safety and efficiency. Simulation results in Gazebo demonstrate significant improvements over a baseline approach, achieving a 95% obstacle avoidance rate and reducing the average encountered slope from 8∘ to 2.7∘ in custom terrain. This integration enhances path efficiency and terrain traversability and supports robust real-time adaptation, paving the way for more reliable autonomous navigation in dynamic outdoor environments.
Figures & tables
Fig. 1: Local path planning with global DEM data integration. (a) The global cost map path generated from SRTM-based elevation data. (b) Local Path Planning Region (30m) extracted from the global map for finer trajectory adjustments. (c) Rviz visualization of real-time LiDAR-based obstacle detection and path correction. (d) Gazebo simulation demonstrating the UGV avoiding obstacles in a dynamic environment (right) compared to the baseline where UGV gets stuck (left)
Fig. 2: (a) Global Processing: Acquisition of DEM data, preprocessing, and generation of the global cost map; (b) Global Path Planning: Optimal trajectory determination using A*; (c) Zoomed-in view of the global cost map, highlighting a selected region for local path refinement; (d) Local Path Correction: Integration of real-time sensor (LiDAR) data for ground segmentation, obstacle detection, and dynamic trajectory adjustment within the selected region.
Reference
OA
SE
RTS
GPO
LPR
[ 1 ]
✓
[ 14 ]
✓
✓
[ 8 ]
✓
[ 15 ]
✓
[ 5 ]
✓
✓
[ 6 ]
✓
TABLE I: Comparison of Our Approach with Related Works
Fig. 3: Visualization of ground (blue points) and non-ground (red points) segmentation from LiDAR data, along with the final path direction (green marker) as visualized in RViz. This image illustrates the result of our local path correction approach, where the UGV dynamically adjusts its trajectory to avoid nearby obstacles and maintain safe navigation through complex terrains.
θcurrent←arccos(∥n∥∥z∥n⋅z)
Algorithm 1 Slope Estimation and Adaptive Traversability Analysis
Fig. 4: Height distribution from terrain in Figure 4(c) as observed by the UGV. The blue histogram shows that most obstacle heights cluster near –0.50 m, with a right tail extending into positive values. The green dashed line indicates the median (–0.45 m) and the red dashed line shows the mean (–0.30 m), confirming a right-skewed distribution ( S>0 ).
Fig. 5: (a) UGV navigating a flat terrain while detecting and avoiding an obstacle (here a pit). The trajectory taken by the UGV is indicated with the blue arrow avoiding the obstacle in its view. (b) UGV traversing a rugged landscape with visible obstacles (pits). The illustration highlights adaptive path planning and obstacle avoidance using our algorithm in action on rough terrain. (c) UGV navigating over a slope, illustrating its capability to handle inclines. The figure depicts how using our algorithm, UGV detects the bump and the adjustments made to avoid and maintain stability.
Fig. 6: Side-by-side comparison of the real Compusult Ltd.’s Nanuk UGV (left) and its ROS model derived from the URDF file (right). This comparison demonstrates the fidelity of our simulation model for sensor integration and navigation evaluation.
Metric
Our Approach
Baseline
Non-Rough
Obstacle
Non-Rough
Obstacle
Avoidance Rate (%)
95
0
Average Slope (°)
0.7
2.7
0.7
8
Time to Goal (s)
43
77
43
Stuck
Path Deviation (m)
0
4.25
0
Stuck
TABLE II: Performance comparison of our approach versus the baseline under varying terrain conditions. NB:* — Not applicable
Robotics Research Group, Indian Institute of Technology · Robotics Research Group, Indian Institute of Technology (BHU), Varanasi, India · University of Colorado, Boulder, USA +2
Centre for Systems and Control, Indian Institute of Technology Bombay, Mumbai · Institute for Control, Optimization, and Networks, Purdue University, West Lafayette, USA