Constraint-Aware Throttle Control for Mission Feasibility Under Tight Energy Budgets

Executive Summary

Energy-constrained autonomous systems fail not because their routes are wrong, but because their energy allocation is. When fuel budgets are tight and disturbances are unpredictable, fixed-throttle policies waste fuel in favorable conditions and starve the actuator in adverse ones.

WHACO (Warranted Hierarchical Autonomous Cooperative Operations) is a mode-switching throttle controller that operates downstream of any existing route planner. It does not modify the route — it modulates how aggressively the agent follows it, selecting from five discrete thrust levels based on local environmental sensing, fuel state, and proximity to objective.

Across 1,000 randomly generated out-of-distribution environments with non-stationary conditions, sensor degradation, actuator lag, model mismatch, and partial observability:

WHACO converts infeasible missions into successful operations. In ultra-tight fuel scenarios where conventional controllers achieve 0% success, WHACO achieves 95% — on the exact same route.

Technical Hypothesis

Hypothesis: State-dependent discrete throttle modulation improves mission feasibility probability under quadratic fuel dynamics relative to fixed-magnitude thrust policies, without requiring modification to directional planning architecture.

Mechanism claim: The improvement derives from real-time burn-rate shaping — reducing thrust when environmental conditions are favorable, increasing thrust when mission geometry demands it, and applying full authority at terminal approach under adverse conditions. This exploits the quadratic relationship between thrust magnitude and fuel consumption, where small reductions in thrust yield disproportionate fuel savings.

Falsification criteria: The hypothesis would be falsified if (a) WHACO's advantage disappeared when tested on identical routes with only throttle magnitude varying (counterfactual test), or (b) random mode selection produced equivalent results to intelligent selection (ablation test). Neither condition was observed.

Technology Readiness

TRL	Description	Status
3	Analytical and experimental proof of concept	Complete — 10-test validation suite, 1,000 OOD worlds, statistical significance established
4	Component validation in laboratory environment	Complete — validated in high-fidelity simulation with non-stationary dynamics, sensor noise, actuator lag, model mismatch, partial observability
5	Component validation in relevant environment	Current — PX4 SITL closed-loop validation (7/7 scenarios); composable swarm autonomy stack with 221 verification checks, 20 certifiable invariants, ASDP survival arbitration with cluster-consistent SVI
6-7	System demonstration in relevant/operational environment	Phase II target — flight testing on UAS platform with real wind conditions, integration with operational MPC planner

Problem

Autonomous platforms operating under energy constraints face a fundamental resource allocation problem: the planner must commit to thrust levels before the full cost of the trajectory is known. When the environment is uncertain and the energy budget is tight, this leads to predictable failure modes:

• Terminal spiraling — low-thrust approaches in headwind zones fail to converge on target
• Premature fuel depletion — high-thrust policies burn reserves during favorable conditions that didn't require them
• Mode collapse — fixed policies cannot adapt to non-stationary environments that shift during mission execution

Fuel depletion follows a quadratic cost model: fuel_used = k · |thrust|² · dt. This creates a direct, non-linear tradeoff between speed and range. A 2x increase in thrust produces a 4x increase in fuel consumption. State-dependent throttle modulation exploits this non-linearity.

Architecture

WHACO operates in a two-stage hybrid pipeline, separating directional planning from energy management:

The route planner provides a unit direction vector. WHACO replaces only the magnitude decision. It requires no modification to the planner and no access to the planning objective function.

Mode-Switching Policy

WHACO selects from eight discrete thrust modes based on real-time sensing:

Mode	Thrust	Trigger Condition
RIDE	20%	Strong tailwind (alignment > 3.0)
CONSERVE	30%	High fuel pressure (fuel_ratio < 0.5)
BRAKE	40%	Look-ahead: headwind increasing
CRUISE	55%	Default — neutral conditions
RECOVER	55%	Post-anomaly stabilization
BOOST	75%	Look-ahead: tailwind increasing
SPRINT	85%	Strong headwind + surplus fuel
PUNCH	100%	Terminal approach or stall override — prevents terminal spiraling

Fuel Pressure Estimation

Compute Overhead

Component	Cost per call	Share of pipeline
MPC direction oracle	836.9 µs	98.2%
WHACO throttle selector	15.1 µs	1.8%

WHACO adds negligible computational overhead. It is suitable for embedded deployment on resource-constrained platforms.

Validation Results

All results are from a ten-test validation suite designed to establish generalization, safety, causal mechanism, and statistical significance. All baselines share the same route planner — only thrust magnitude varies.

Out-of-Distribution Generalization (1,000 worlds)

Each world has independently randomized start/goal positions, wind structure, fuel budget, and five categories of perturbation:

Metric	MPC Baseline	WHACO	Delta
Mission success rate	86.7%	96.5%	+9.8pp
Mean fuel consumed	4.489 units	3.736 units	-17.4%
Missions rescued (WHACO wins)	—	105
Missions lost (WHACO loses)	—	7
Rescue-to-loss ratio	—	15:1

Perturbation Robustness

Perturbation Type	N Worlds	WHACO Win Rate
Non-stationary wind (corridor drift)	487	84.0%
Burst sensor noise (spikes)	486	82.7%
Actuator lag (1-3 timestep delay)	487	88.9%
Model mismatch (drag 0.5x-1.5x)	591	82.9%
Partial observability	430	81.9%

WHACO's advantage is stable across all perturbation categories. It is not exploiting a specific environmental feature.

Counterfactual Fairness (Same Route, Different Throttle)

All policies use the same live MPC direction oracle called from each agent's own position, with identical wind noise seeds. Only thrust magnitude varies.

Scenario	MPC 75%	WHACO	Constant 60%	Bang-Bang
Tight fuel (5.0)	44%	100%	100%	91%
Moderate fuel (8.0)	100%	100%	100%	64%
Hostile headwind	100%	100%	100%	68%
Ultra-tight (4.0)	0%	95%	72%	89%

Statistical Significance (Bootstrap, n = 10,000)

Metric	Point Estimate	95% Bootstrap CI	Effect Size
Success delta	+9.8pp	[+7.8, +11.8]pp	CI excludes zero
Fuel savings	+17.4%	[+16.5, +18.4]%	Cohen's d = 1.30
Time cost	+17.6%	[+17.0, +18.3]%	Cohen's d = 1.65
Paired t-statistic	38.0	df = 859	p < 0.01
Rescue ratio	105:7 (15:1)	[87-124]:[2-13]	CIs non-overlapping

Worst-Case Safety

Adversarial Scenario	MPC Success	WHACO Success	Delta
Trap corridor (extreme strength)	100%	100%	+0pp
Zero structure (no corridor)	100%	100%	+0pp
High sensor noise (σ = 4.0)	100%	100%	+0pp
Tight headwind (fuel = 5)	0%	86%	+86pp
Reverse corridor (away from goal)	100%	100%	+0pp
Ultra-tight fuel (fuel = 4)	0%	100%	+100pp

No degradation in mission success was observed across tested adversarial conditions. In the two hardest cases, it rescues missions that are infeasible under conventional throttle policies.

Ablation Analysis

Systematic removal of each WHACO mechanism identifies the causal contribution of each subsystem:

Variant	Avg Fuel Savings	Avg ΔSuccess	Tight Fuel Success
WHACO (full system)	+12.9%	+10.5pp	100%
No look-ahead probe	+15.4%	+10.5pp	100%
No terminal mode	+28.5%	+10.5pp	100%
Flat (always 55%)	+33.4%	+10.5pp	100%
Random mode selection	+10.7%	+7.5pp	88%

Integration Specification

Interface

Deployment Gating Rule

Operational Envelope

Constraint Level	WHACO Value	Recommendation
Tight fuel (< 2x minimum)	+56 to +95pp success	Always enable — mission-critical
Moderate fuel	12-18% fuel savings	Enable for operational efficiency
Loose fuel (> 3x minimum)	7-10% fuel savings	Optional — marginal benefit
Extreme noise (σ > 4.0)	Untested	Disable — fail-safe

Composable Autonomy Stack (v2.0)

Beyond single-agent throttle control, WHACO v2.0 extends to a composable multi-agent autonomy stack for swarm-scale operations. The stack comprises five authority layers with strict precedence and deterministic conflict resolution.

Layer	Module	Authority	Capability
L5	Resilience	Highest — safety-critical	ASDP survival arbitration, degradation, SCORCH technology denial
L4	Node Switching	Continuity	Heartbeat monitor, hot standby, 1-tick handoff
L3	Governance	Consensus	Election, posture quorum, task allocation, merge reconciliation
L2	Predictor	Sensing	Wind estimation, fuel forecast, threat prediction, anomaly detection
L1	Base Governor	Execution	11-mode throttle policy, 15µs/call

Adaptive Survival & Denial Protocol (ASDP)

ASDP is a survival-first arbitration layer within Resilience (L5). It computes a Swarm Viability Index (SVI) per connected cluster and exhausts non-destructive survival modes before falling through to technology denial:

Cluster-consistent SVI: During network partition, SVI is computed per connected component. Each cluster selects its own ASDP mode independently. A healthy cluster stays NOMINAL while a dying cluster escalates to denial.

Anti-oscillation: Asymmetric hysteresis prevents mode thrashing. Escalation is immediate. De-escalation requires SVI to exceed threshold by +0.05 margin, a 10-tick lockout, and single-step progression.

Verification at Scale

Suite	Checks	Coverage
ASDP Validation	82/82	SVI, mode selection, hysteresis, cluster SVI, R0 authority
SCORCH Lifecycle	70/70	5-condition AND gate, interlocks, veto, 4 modes
Authority Matrix	15/15	All pairwise module conflicts including R0
N=100 Collapse	50/50	Full-composition deterministic collapse
Total	221/221	20 certifiable invariants, < 7.5ms compute @ N=100

Hardware Transition Plan

The primary transition target is a PX4-based UAS testbed operating in outdoor wind conditions. The integration architecture preserves WHACO's bolt-on design principle: the controller inserts downstream of the existing autopilot's MPC direction output, modulating only thrust magnitude.

Phase I: PX4 SITL Closed-Loop Validation COMPLETE

WHACO has been validated inside PX4 Autopilot v1.14.3 running in SITL mode with jMAVSim. The companion process connects via MAVLink (UDP 14540), reads vehicle state at 50Hz, runs mode selection, and issues OFFBOARD velocity commands. 7/7 scenarios pass across fuel-stress, terminal headwind, and extended mode-cycling conditions.

Component	Specification	Status
Autopilot	PX4 v1.14.3 (SITL + jMAVSim)	Validated
Insertion point	OFFBOARD mode velocity commands scaled by WHACO thrust fraction	Validated
Wind sensing	Groundspeed-based wind proxy (EKF2 WIND_COV unavailable in jMAVSim); stall detection override	Validated
Fuel/energy state	PX4 SYS_STATUS battery percentage → fuel model mapping	Validated
Compute platform	Companion process via MAVLink UDP. 50Hz control loop with 10Hz telemetry capture.	Validated
Telemetry	CSV + JSON: mode transitions, fuel pressure, wind alignment, battery, position, PX4 nav state	Validated
State management	OFFBOARD → MANUAL → force-disarm → heartbeat confirmation; arm retry with 5 attempts	Validated

PX4 SITL Results

Scenario / Policy	Result	Dist	Time	Battery
A / WHACO	PASS	3.6m	21.0s	62%
A / Fixed 75%	PASS	3.5m	15.0s	63%
B / WHACO	PASS	3.6m	3.3s	72%
B / Fixed 55%	PASS	3.7m	3.0s	67%
B / Fixed 75%	PASS	3.7m	2.7s	67%
III / WHACO	PASS	3.7m	20.0s	62%
III / Fixed 75%	PASS	3.8m	15.6s	70%

Phase II: Flight Test

Outdoor flight testing under real wind conditions (5-15 kt sustained, gusts to 25 kt). Test protocol: identical waypoint missions flown back-to-back with WHACO enabled vs. fixed-throttle baseline. Metrics: battery consumption delta, mission completion rate under deliberately constrained battery, terminal approach accuracy in crosswind.

Application Domains

The core capability — maintaining mission feasibility under tight energy budgets in uncertain environments — maps to multiple autonomous platform categories:

Domain	Disturbance Analog	Energy Analog	Value Proposition
Defense swarms (sUAS)	Threats, EW, attrition	Battery + survivability	Survivable multi-agent autonomy with ASDP, N=100+ scale
Delivery drones (UAS)	Wind, thermals	Battery	Rescue battery-critical last-mile deliveries
Electric vehicle fleets	Grade, temperature, traffic	Battery range	Reduce range anxiety, extend serviceable area
Maritime autonomy (USV)	Ocean currents, tides	Fuel reserves	Current-aware throttle for operational range
Warehouse AMRs	Congestion, queue delays	Battery charge	Charge-aware throughput optimization
Aircraft operations	Jet stream, turbulence	Fuel burn	Fuel burn shaping for marginal-range flights
Spacecraft	Gravity wells, solar wind	Propellant	Propellant-optimal trajectory following

Limitations

1. Time cost. WHACO trades time for fuel (+17.6% mean mission time). For time-critical missions with adequate fuel, standard throttle may be preferred.
2. Empirical validation, not formal proof. Safety properties were observed across all tested scenarios but have not been formally verified. Safety-critical deployment would require additional formal analysis and certification.
3. 2D domain validation. Current test environment is a 2D continuous plane. Transfer to 3D, turbulent, or discontinuous environments requires further validation.
4. Fixed mode thresholds. Mode selection uses calibrated thresholds that may require domain-specific tuning for different physics models.
5. SITL only. PX4 validation is software-in-the-loop; outdoor flight testing under real wind conditions is Phase II.

About Blackridge Autonomy

Blackridge Autonomy LLC develops mission-critical adaptive control systems for energy-constrained autonomous platforms operating in degraded, energy-constrained, and contested operational environments.

Our technology is designed as a bolt-on control layer — deployable downstream of existing planners with no system rearchitecture required. We focus on the energy management problem: maintaining mission feasibility when fuel budgets are tight and environments are uncertain.

NAICS Codes

• 541715 — Research and Development in the Physical, Engineering, and Life Sciences
• 541330 — Engineering Services
• 541511 — Custom Computer Programming Services
• 541512 — Computer Systems Design Services

Contact

This document is UNCLASSIFIED and approved for public release. Distribution Statement A: Approved for public release; distribution is unlimited.