Systems OptimisationSimulationPython MATLABRoboticsEnergy Systems University of Southampton

MFC-Powered Robotic Arm
Systems Optimisation

A four-layer systems optimisation framework establishing the mechanical, energetic, and electrochemical requirements of a three-link robotic arm powered entirely by Microbial Fuel Cells — harvesting energy from organic matter in its operating environment to sustain fully autonomous, indefinite operation without human intervention.

Year 2026 Word Count 10,471 Tools Python · MATLAB Simulation Outputs ~30 figures

2,461

Theoretical actuations from 10g apple

2.26J

Energy per cycle — EMAX locked-out

28cm²

Min. electrode area for 1 action/day

MFCs in series minimum

Optimisation layers in framework

Overview

The problem

Fully autonomous robots must sustain themselves indefinitely without human intervention — but conventional battery-powered systems impose a fundamental limit on true autonomy, requiring periodic recharging or battery replacement.

This project addressed that constraint directly: can a low-power robotic arm be powered entirely by Microbial Fuel Cells, harvesting energy from organic matter in its operating environment — fallen fruit, fly biomass, organic waste — with no external power input?

Two real-world deployment scenarios framed the work. An orchard rover collecting fallen fruit and depositing it into an artificial stomach generating power through MFCs. A forest survey robot sustaining itself on insect biomass in a remote environment unsuitable for long-term human presence.

Approach

Four-layer framework

Rather than optimising components in isolation, a layered, staged framework was developed — each layer producing defined inputs and outputs for the next, ensuring every design decision was grounded in the physical constraints of the full system.

Mechanical Feasibility

Screen actuators against the shoulder torque constraint. Eliminate any that cannot physically drive the arm geometry.

Energy Minimisation

Model energy per cycle for each feasible actuator across motion and holding phases using the Operating Current model.

Storage & Charging

Size the supercapacitor and determine charging time. Compare battery vs supercapacitor for long-term autonomous viability.

MFC Stack Design

Establish minimum cell count, series/parallel configuration, and electrode area to meet power delivery requirements.

Layer 1 — Mechanical Feasibility

Actuator selection

Six off-the-shelf actuators were evaluated against a single governing constraint: could each actuator overcome the shoulder joint torque at worst-case loading — arm fully horizontal, carrying the payload and the weight of all distal actuators? Three were eliminated before energy analysis began.

Actuator	Mass (g)	τ Required (Nm)	τ Available (Nm)	Feasible	Note
EMAX ES08A II	8	0.0781	0.120	✓ Yes	Optimal selection
N20 100:1	10	0.0831	0.150	✓ Yes	Self-locking gearbox
SG90	9	0.0806	0.180	✓ Yes	Feasible, higher energy draw
DS3218	60	0.2082	0.190	✗ No	Self-defeating — mass exceeds own capability
Pololu 30:1	10	0.0831	0.080	✗ No	Marginally below stall torque
28BYJ-48	30	0.1332	0.034	✗ No	Greatly exceeds stall torque

Locked out vs always powered bar chart — Fig. 1 — Locked-out vs always-powered energy comparison (I_op model, 60s reference cycle). The EMAX ES08A II shows the greatest benefit from locked-out operation.

Results

Key findings

⚡

Energy per cycle

The EMAX ES08A II was the most energy-efficient actuator in locked-out mode. The locked-out strategy — where actuators draw no power during charging — proved critical to system viability, with the always-powered case increasing EMAX consumption by 3× over a 19-hour charge.

2.26J/cycle — EMAX locked-out

🔋

Dominant energy phases

Phase 2 (lift and retract) and Phase 6 (lower arm) together account for approximately 60% of total motion energy, driven by the simultaneous activation of three joints. Reducing sweep angles and increasing actuation speed are the highest-leverage optimisation targets.

~60% of energy in 2 of 6 phases

🍎

Substrate analysis

10g of fresh apple flesh provides 5,550J of extractable electrical energy at 25% Coulombic efficiency — theoretically supporting over 2,000 actuations. Fly biomass offers 2.2× more extractable energy per gram, though its MFC power density remains uncharacterised.

5,550J from 10g apple | 2.2× from fly biomass

📈

Electrode area — the key lever

Electrode area was identified as the primary optimisation target, with a linear correlation to daily actuation frequency. For the EMAX to achieve one action per day using fresh apple flesh, a minimum electrode area of 28cm² is required — approximately the size of a credit card.

28cm² for 1 action/day — credit card scale

🔌

MFC stack configuration

A minimum of two MFC cells in series is required to meet the 0.25V cold-start threshold of the LTC3105 DC-DC boost converter. A supercapacitor (0.69–0.73F) was confirmed as optimal over a battery, offering 500,000+ cycles versus 500–1,000 for lithium-ion.

2 cells minimum | 500,000+ supercapacitor cycles

⚙️

Duty cycle

Using fresh apple flesh, the EMAX achieves a duty cycle of approximately 0.0279% — 5 seconds of motion per ~18,000 seconds of charging. With lab-optimised acetate as fuel, this improves to 0.881%, demonstrating the system is energy self-sufficient and power-limited, not energy-limited.

0.0279% duty cycle — apple flesh

Simulation & Modelling

Simulation & modelling

All simulations were conducted in Python across three iteratively developed scripts, producing approximately 30 output figures. The framework covered actuator energy modelling, MFC stack configuration, and energy storage sizing.

The Operating Current model was selected over the efficiency model, as it draws directly from manufacturer datasheets — providing actuator-specific values without uniform efficiency assumptions that would mask differences between geared motors, servos, and stepper motors.

MATLAB was used to visualise the 3D arm geometry and validate joint configurations across both motion scenarios: a pick-and-place claw effector (Scenario A) and a base-rotation scoop effector (Scenario B).

All results are deliberately conservative, using worst-case torque assumptions and upper-bound operating currents — ensuring values represent a safe minimum for real-world application.

PythonMATLAB MatplotlibNumPy Sensitivity analysisTornado charts Datasheet analysis

Conclusion

System feasibility

MFC-powered robotic actuation is systems-feasible

The objectives of the project were met. A viable off-the-shelf actuator was identified, its energy use quantified, and the full power chain from organic fuel to mechanical motion was established within realistic physical and economic constraints.

Energy feasibility satisfied for all substrates — 10g of apple flesh supports 2,461 theoretical actuations at the EMAX configuration
Power feasibility is conditional on electrode area — 28cm² achieves 1 action/day for the EMAX using fresh apple flesh
The EMAX ES08A II with active power management (locked-out) is the recommended actuator for real-world deployment
A two-cell series MFC stack with LTC3105 converter and 0.69–0.73F supercapacitor forms the minimum viable system
Electrode area is the primary lever for optimisation — a linear relationship between area and daily actuation frequency

Fig. 4 — MPP voltage vs internal resistance. Fresh apple flesh (0.2V) falls below the LTC3105 minimum threshold. Two cells in series (0.4V) resolves this.

MATLAB simulation output — Scenario A, Phase 3: Yaw to MFC. Left: 3D arm configuration. Centre: 2D side profile with joint angles. Right: real-time cumulative energy accumulation comparing baseline (2.25J) vs optimised (2.21J, −1.8%).

Simulation Recording

MATLAB motion simulation

Animated MATLAB simulation of Scenario A — the full pick-and-place motion cycle across all six phases, from initial position through payload collection, yaw to the MFC deposit zone, and return.

Scenario A motion cycle — EMAX ES08A II configuration. Six phases: close claw → lift & retract → yaw to MFC → open claw → reverse yaw → lower arm. Total motion time: 5 seconds.

System Feasibility

Feasibility heatmap

The heatmap shows daily actuation frequency across all feasible actuators, fuel types, and MFC cell counts for both Scenario A and B. Green indicates high actuation frequency; red indicates near the feasibility limit. The EMAX with acetate achieves the highest performance across both scenarios.

Recommendations

Further work

The simulation framework establishes a solid theoretical foundation. The following experimental steps would convert model predictions into validated, deployable design data.

Fresh apple flesh MFC characterisation

Directly measure the power density and open-circuit voltage of a fresh apple flesh MFC to replace the estimated 20mW/m² with an experimentally validated value.

Fly biomass MFC characterisation

Measure MFC power density of fly biomass across preparation methods. Experimental data would resolve the sensitivity uncertainty — the single most impactful further work recommendation.

Actuator characterisation at operating point

Measure actual current draw at the specific torques and speeds of this arm geometry to validate the Operating Current model against physical hardware.

Inverse kinematics optimisation

Implement inverse kinematics to identify minimum-energy joint angle trajectories for each motion scenario — reducing total energy without any hardware changes.

Regenerative braking assessment

Assess energy recoverable from gravity-assisted phases — specifically elbow counterrotation (Ph2) and shoulder lowering (Ph6) — to evaluate feasibility at this scale.

Physical prototype & validation

Construct and test the proposed system experimentally — measuring total energy, charging time, and actuation frequency across both substrates.

MFC-Powered Robotic ArmSystems Optimisation