Source-linked AI summary
Fly, Drive, Reconfigure: A Modular Reconfigurable Aerial-Ground Platform for Field Operations
Li-Yu Lo, Yanbaihui Liu, Chengchuan Shu, Tyler Harris, Jonathan Ryan, Boyuan Chen
TL;DR
Fixed robot roles and the contrasting energy and terrain constraints of flight and ground travel limit field missions requiring broad sensing, traversal, and payload deployment. HARP combines independently deployable scouts, rovers, and payloads with terrain-aware air-ground planning and autonomous reconfiguration. Outdoor experiments demonstrate terrain-dependent mobility selection, reconfiguration benefits, and end-to-end air-ground operation with a sampling payload.
Problem
Field robots face trade-offs among sensing, mobility, payload capacity, and endurance, while physical reconfiguration has not been integrated broadly with perception, mobility planning, heterogeneous roles, and mission execution.
Source: §I · p. 1Method
HARP uses scouts to map terrain for energy-aware route and mobility planning, while independently flying rover and payload modules autonomously assemble for ground transport.
Source: §I · pp. 1–2Results
Outdoor experiments demonstrate terrain-dependent mobility selection, energy and payload benefits of reconfiguration, autonomous assembly, and end-to-end air-ground operation with a sampling payload.
Source: §V · p. 7Takeaways & Limitations
HARP demonstrates that heterogeneous teams can adapt not only their actions but also how their physical capabilities are composed during a mission.
Source: §V · p. 7Longer-duration missions under extreme wind, uneven terrain, and intermittent communication remain future evaluation directions.
Source: §V · p. 7
Abstract
from arXiv · showhide
Heterogeneous robot teams distribute complementary capabilities across specialized agents, but their physical roles and capacities typically remain fixed throughout a mission. We present HARP, a Heterogeneous Aerial Robotic modules Platform in which independently deployable aerial robots physically reconfigure to compose their capabilities for field operations. HARP comprises sensor-equipped scouts, flydrive rover modules, and task-specific payload modules. Scouts map the environment and inform an energy-aware planner that jointly selects routes and air-ground mobility modes. Rover and payload modules fly independently across terrain that constrains ground travel, then autonomously assemble into a cooperative ground vehicle for energy-efficient payload transport. Motivated by environmental sampling in remote and difficult-to-traverse regions, we evaluate HARP through field experiments spanning sensing, planning, reconfiguration, airground mobility, payload transport, and task execution. We further conduct module-level deployment tests on the Greenland Ice Sheet toward future autonomous missions. HARP demonstrates how heterogeneous robot teams can adapt not only their actions, but also how their physical capabilities are composed during a mission.
I. INTRODUCTION
Conventional robots trade off sensing, mobility, payload capacity, and endurance, while heterogeneous teams generally coordinate actions without physically recomposing their capabilities. HARP combines independently deployable aerial modules with autonomous reconfiguration and terrain-aware air-ground planning for field operations.
Motivation
- Aerial robots traverse challenging terrain quickly but use substantial energy, while ground robots travel efficiently and carry heavier payloads but remain terrain-constrained.These trade-offs are especially limiting in missions requiring large-area sensing, long-distance traversal, and task-specific payload deployment.Source: §I · p. 1
- Heterogeneous teams distribute complementary roles, but fixed physical capacities leave missions needing more actuation or payload capacity than one robot can provide.Modular reconfiguration has often been studied as a robot capability rather than as part of an integrated field-autonomy pipeline,,,,.Source: §I · p. 1
HARP platform
- HARP combines sensor-equipped scouts, fly-drive rovers, and task-specific payloads that deploy independently by air and can assemble into a cooperative ground vehicle.Multiple rovers compose their capabilities for energy-efficient payload transport while retaining independent flight when aerial mobility is advantageous.Source: §I · pp. 1–2
Autonomy and coordination
- A scout maps terrain for an energy-aware planner that jointly selects routes and air-ground modes, while rover-payload modules transition between formation flight and assembled ground traversal.Autonomous docking and disassembly support physical capability composition during a mission.Source: §I · p. 2
Evaluation
- Outdoor experiments cover sensing, planning, reconfiguration, payload transport, and integrated sampling missions, alongside preliminary module-level tests on the Greenland Ice Sheet.The Greenland tests inform future field deployments.Source: §I · p. 2
II. RELATED WORK
Prior heterogeneous teams coordinate sensing, mobility, and task roles, while modular systems physically combine robots to expand capabilities. HARP brings physical reconfiguration into heterogeneous air-ground field missions.
Heterogeneous Air-Ground Robotic Systems
- Heterogeneous field teams have demonstrated mapping, exploration, inspection, and long-duration autonomy, but their physical embodiments and carrier-payload relationships generally remain fixed.Earlier work emphasizes state estimation, communication, planning, and coordination,,,,,, [28].Source: §II · p. 2
Modular and Reconfigurable Robotic Systems
- Modular robots connect units to form larger systems, with work spanning modular architectures and increasingly deployable, perception-driven operation,,,,,,,,.Modules have also formed flying structures, while multimodal and cooperative-transport systems extend mobility, actuation, and payload capabilities,,,, [40],,.Source: §II · pp. 2–3
- HARP uses independently deployable aerial modules that assemble into a cooperative ground vehicle when collective ground mobility and payload capacity are advantageous.This brings physical composition into a heterogeneous air-ground mission rather than changing only how robots share information and tasks.Source: §II · pp. 2–3
III. METHOD
HARP integrates scout-based terrain mapping, energy-aware route and mobility planning, independently flying modules, and autonomous assembly into a cooperative ground vehicle. Its planner represents terrain and mobility modes to choose feasible air-ground routes.
A. System Overview
- A scout maps terrain with LiDAR-SLAM and GPS; a centralized planner selects route segments for assembled ground traversal or independent flight, with docking and disassembly enabling mode changes.A group-level rigid-body controller maps assembled-system motion to individual rover commands.Source: A. System Overview · p. 3
B. Module and Mechanical Design
- Four rover modules attach to the payload quadrotor through electromagnetic interfaces, forming a symmetric ground vehicle with passive compliance at each connection.Springs, a pitch hinge, and axial clearance help maintain track contact on uneven terrain without additional suspension actuators.Source: B. Module and Mechanical Design · p. 3
C. Mapping and Air-Ground Path Planner
- LiDAR points form a 2.5-D elevation-and-roughness map with obstacle and unsafe-landing masks; an eight-connected planner jointly selects ground/flight routes using feasibility and energy costs.Unobserved cells permit neither ground travel nor landing; reverse Dijkstra search from ground-mode targets produces the cost-to-go map and minimum-energy route.Source: C. Mapping and Air-Ground Path Planner · pp. 3–4
D. Formation Flight
During aerial segments, four rovers follow the payload in formation using a leader-rooted consensus structure that preserves docking offsets for later transitions.
D. Formation Flight
- For aerial path segments, the payload acts as formation leader while N = 4 rover modules fly independently at prescribed offsets, placing them near required docking geometry at the subsequent transition site.Source: D. Formation Flight · p. 4
- The formation controller represents the payload and rovers as nodes in a directed leader-rooted graph, where edges specify whose state followers receive.The graph contains a directed spanning tree rooted at the payload.Source: D. Formation Flight · p. 4
- Zero position-consensus error corresponds to each rover holding its desired offset from the payload, pi = p0 + ri.The desired offsets are constant vectors in three-dimensional Cartesian space.Source: D. Formation Flight · p. 4
E. Vertical Docking and Reconfiguration
At ground-transition sites, the rovers align and dock the payload; the assembled modules then coordinate ground motion as a planar rigid body.
E. Vertical Docking and Reconfiguration
- Rover 1 leads docking alignment while three followers regulate offsets; the payload descends to the converged formation center, where four electromagnetic locks secure it.Rover 1 aligns its yaw with the planned ground-path heading; disassembly reverses the sequence, releasing the locks and returning the modules to independent flight.Source: E. Vertical Docking and Reconfiguration · p. 4
F. Coordinated Ground Motion
- After docking, the four rovers and payload are modeled as one planar rigid body; attachment vectors map payload motion, including rotation-induced tangential velocity, to rover velocities.All modules share the desired yaw rate.Source: F. Coordinated Ground Motion · p. 4
- Velocity allocation assumes fixed attachment geometry and omits passive connector compliance and track-terrain slip.Source: F. Coordinated Ground Motion · p. 4
A. Terrain Mapping and Air–Ground Planning
Scout-reconstructed terrain guides HARP’s air–ground mode choices, while physical reconfiguration reduces modeled pre-task energy and increases assembled tractive force. The planner switches between ground and aerial segments where mapped terrain makes ground travel infeasible.
A. Terrain Mapping and Air–Ground Planning
- On scout-mapped terrain, the planner kept all three routes in assembled ground mode when access was continuous, but added aerial segments around obstacles and elevation discontinuities.The routes returned to ground operation near the goal rather than following a fixed sequence of mobility modes.Source: A. Terrain Mapping and Air–Ground… · p. 4, Figure 3 · p. 5
B. Benefits of Physical Reconfiguration
- 4.18 Wh mean pre-task energy for HARP versus 5.73 Wh for the monolithic platform and 6.10 Wh for aerial-only, across three initial conditions.These correspond to reductions of 27.1% and 31.5%, respectively; the ground-only baseline had no feasible route under the terrain and collision constraints.Source: B. Benefits of Physical Reconfiguration · p. 5, Figure 4 · p. 5
- 87.6 N for the docked four-rover configuration versus 22.4 N for one rover, a 3.91× increase in measured tractive force.Force was measured on a controlled mat surface with motors at full actuation.Source: B. Benefits of Physical Reconfiguration · p. 5, Figure 4 · p. 5
C. Autonomous Docking Performance
Repeated trials measured autonomous docking reliability, followed by an outdoor demonstration of ground–air transitions and drilling. The demonstration achieved a 90.75% autonomy ratio, while the drill’s sampling performance remained limited by soil conditions.
C. Autonomous Docking Performance
- Nine of 11 vertical-docking trials succeeded (81.8%); mean touchdown error was 0.11 ± 0.04 m, with a maximum of 0.17 m.Success required touchdown within 0.15 m of the rover-formation center and successful engagement of the magnetic lock.Source: C. Autonomous Docking Performance · pp. 5–6
D. End-to-End Outdoor Field Demonstration
- The 8×18 m outdoor course supported a 26.12 m mission with two mobility transitions, a 90.75% autonomy ratio, and human intervention only for final formation adjustment before docking.The route included 3.42 m in ground mode and 22.7 m in aerial mode.Source: D. End-to-End Outdoor Field… · p. 6, Figure 6 · p. 6
- The drill collected soil from humid ground, but could not reliably extract a complete specimen or penetrate harder soil.A more capable drill was under development for future field testing and firn sampling.Source: D. End-to-End Outdoor Field… · pp. 6–7
E. Module-Level Field Tests on the Greenland Ice Sheet
Outdoor experiments demonstrate HARP’s terrain-dependent mobility selection, reconfiguration benefits, autonomous assembly, and end-to-end sampling operation. Greenland tests assessed individual-module flight and rover traversal to inform future deployments.
E. Module-Level Field Tests on the Greenland Ice Sheet
- All modules completed individual flight tests in Greenland, and rover modules also traversed local terrain to inform future deployment requirements.The tests assessed deployment logistics and module-level operation outside the primary experimental setting.Source: E. Module-Level Field Tests on the… · p. 7
V. CONCLUSIONS, LIMITATIONS, AND FUTURE WORK
- Outdoor experiments demonstrate terrain-dependent mobility selection, energy and payload benefits of reconfiguration, autonomous assembly, and end-to-end air-ground operation with a sampling payload.These results cover integrated system operation as well as physical reconfiguration.Source: §V · p. 7
- Autonomous docking succeeded in 9 of 11 trials, while drilling stability for sampling on harder surfaces remains to be evaluated.The drilling experiment demonstrated integrated task execution in soft soil.Source: §V · p. 7
- Future work targets more robust reconfiguration and evaluation over longer missions with extreme wind, uneven terrain, and intermittent communication.Proposed robustness measures include relative-pose feedback, compliant capture mechanisms, and mechanically retained weather-resistant connectors.Source: §V · p. 7
APPENDIX SUPPLEMENTARY METHOD DETAILS
The planning cost model expresses edge costs in watt-hours, accounts for rover and payload mass, and computes a minimum-cost route through a hybrid graph.
A. Air–Ground Planning Cost Model
- Edge costs are expressed in Wh, with payload mass shared across four rovers so each carries its own mass plus one quarter of the payload mass.The ground-cost model also uses path length, roughness, grade, and speed as inputs.Source: A. Air–Ground Planning Cost Model · p. 7
- An aerial edge costs [4Pcr(mr) + Pcr(mp)]d/(3600va), with transition energies added for takeoff, landing, docking, and undocking.Source: A. Air–Ground Planning Cost Model · p. 7
- Reverse Dijkstra search from the goal computes cost-to-go values on the hybrid graph, and routes follow the minimizing successor.The cost-to-go recurrence minimizes the edge cost plus the successor’s cost-to-go.Source: A. Air–Ground Planning Cost Model · p. 7
B. Formation-Control Derivation
The formation-control derivation defines a leader-pinned graph matrix and gives each follower a proportional-derivative control law for constant offsets.
B. Formation-Control Derivation
- The follower adjacency matrix A, degree matrix D, and leader-pinning matrix B define H = D − A + B.Source: B. Formation-Control Derivation · p. 7
- For a leader-rooted graph, H is nonsingular.Source: B. Formation-Control Derivation · p. 7
- For constant offsets, each follower applies u_i = a_0 − K_p e_i − K_d ė_i, with positive gains K_p and K_d.Source: B. Formation-Control Derivation · p. 7