Adding a UGV teammate can slow the mission down
The paper's central warning is that heterogeneous teaming does not universally improve performance. A naive human-UGV strategy, which offloads casualty carrying to the robot whenever possible without optimizing where the handoff happens, increased mean mission time by 2.0% relative to the human-only baseline. The penalty was most pronounced in the study's Cases 2 and 3, where dense terrain forced the UGV to travel substantially longer distances than the direct human route, negating its speed advantage.
The mechanism is asymmetric traversability. The UGV's traversable subgraph is a strict subset of the human's: the human can pass shallow water up to 1.2 m deep and dense forest, while the Warthog cannot. Slopes above 45% unloaded and above 20% loaded are impassable for both. So the robot's speed on open terrain and roads only pays off if the handoff points are placed where each agent is actually operating inside its own traversability envelope.
Human and UGV costs are not measured in the same units
The human agent is modeled with the Pandolf equation, which predicts metabolic rate in watts from body mass, external load, terrain factor, walking speed, and percent grade. A second stage applies the Santee downhill correction and a three-tank hydraulic analogue of energy reserves, so walking speed is modulated by cumulative fatigue: the human slows progressively over long or strenuous segments and recovers partially on easier terrain.
The UGV uses a kinematic model instead. Edge energy comes from rolling resistance, F = mg(sin θ + C_rr cos θ), with vehicle mass of 280 kg unloaded and 380 kg when carrying the 100 kg casualty, and C_rr assigned by surface type. Effective speed is the nominal land-cover speed scaled by max(0.1, 1 − |g%|/g_max), with g_max of 45% unloaded and 20% loaded. The UGV experiences no fatigue.
Because neither model produces costs in shared units, the framework does not sum edge costs across agents. It optimizes mission-level objectives, primarily total time. This is the design decision that makes the switch-point problem meaningful rather than a simple shortest-path join.
Switch points decide who carries the casualty where
A switch point is a casualty transfer location between the human team and the UGV. In the representative scenario the paper describes, the optimized strategy assigns the human to carry the casualty through a short region impassable to the UGV while the UGV simultaneously routes around the obstacle. Concurrent execution of these roles is what produces the time savings over both baselines.
The optimization solves for all three switch points jointly, including the initial human-UGV rendezvous. That matters because it effectively optimizes the UGV's goal location too, letting the planner find a rendezvous that shortens the human carry distance and total mission time at the same time. Switch points are constrained so none lies in an area inaccessible to the UGV, and the problem is solved with mixed discrete particle swarm optimization with diversity preservation, run for 100 iterations with a population of 100 particles.
Faster missions cost more human energy than the naive plan
Both human-UGV strategies cut human energy expenditure relative to the human-only baseline, but the naive strategy cut more: 22.4% versus 17.8% for the optimized strategy. That is expected, because the naive strategy maximizes UGV loaded time regardless of routing cost, offloading the most effort from the human at the expense of a longer overall mission.
The optimized strategy accepts a slightly smaller energy reduction in exchange for consistent time savings. The paper frames this as the right trade in time-critical casualty evacuation, where minutes can determine patient outcome. Readers should keep the two metrics separate: mission time is measured against the human-only baseline and the naive human-UGV strategy, while human energy expenditure is a percentage change against the human-only baseline only.
Case 3 is the exception worth noting. There the optimized strategy outperformed the naive approach on both mission time and human energy expenditure simultaneously, which the authors attribute to jointly optimizing all three switch points rather than fixing the rendezvous.
Short 1 km2 missions understate fatigue and battery limits
The evaluation used multiple synthetic 1 km2 environments with procedurally generated elevation and land-cover data at 2 m per cell, a resolution chosen partly because public digital elevation and land-cover maps are limited to 30 m and partly to accommodate the Warthog's 1.38 m width so each agent occupies a single cell. Mission duration was kept below the platform's nominal 2.5 hour runtime.
The authors state that this scale limits the expression of two key cost drivers: human fatigue and UGV battery consumption. In longer missions, fatigue-modulated speed degradation would increasingly favor UGV casualty transfer, and energy constraints on both agents would raise the stakes of suboptimal switch point placement. The paper expects the performance gap between naive and optimized strategies to widen with mission scale, but that expectation is stated as an expectation, not a measured result.
A second boundary is the planning-time cost model itself. The rolling-resistance formulation operates on categorical land-cover data. Execution-time models such as the Stribeck friction-based cost function in AnyNav use per-cell friction prediction from visual input, roughness, resistance, and turning penalties, inputs unavailable in a synthetic environment with procedurally generated terrain. In deployment, the categorical C_rr and nominal speed assignments would be replaced by continuously predicted friction and speed profiles from a visual perception model.
Coordination, not robot count, is what buys the time
The decision-relevant comparison in this paper is not human versus robot but naive teaming versus coordinated teaming. Deploying a UGV without optimizing transfer locations produced a 2.0% mean mission time penalty against the human-only baseline, while the same hardware with optimized switch points produced a 5.3% reduction against that baseline and a 7.0% reduction against the naive strategy. The optimized strategy achieved the shortest mission time in every scenario tested.
For readers weighing human-UGV casualty evacuation, the paper's own variables point to a concrete priority: solve for switch point locations, including the initial rendezvous, whenever the human and robot have non-comparable cost structures and asymmetric traversability. The energy-versus-time trade-off is the second-order decision, and the paper's stated preference is to accept the smaller 17.8% energy reduction in exchange for consistent time savings in time-critical evacuation.
The scenario split to watch is terrain density. The naive strategy's penalty concentrated in Cases 2 and 3, where dense terrain forced long UGV detours. That is where switch point optimization earns its keep, and where the paper's own reasoning suggests the advantage would grow as missions lengthen and fatigue and battery limits begin to bind.
Sources used in this answer
Traversability-Aware Cooperative Path Planning for Human-UGV Casualty Evacuation
Optimizing casualty transfer switch points for a two-agent human-UGV team reduced mean mission time by 5.3% versus a human-only baseline and 7.0% versus a naive human-UGV strategy, while cutting human energy expenditure by 17.8%; the naive strategy cut energy more (22.4%) but increased mission time by 2.0%.