Economic Mosaic is proud to announce that Amshul Medisetti is the winner of the $6,500 Economic Mosaic STEM Research Scholarship. Congratulations, Amshul.

The scholarship was open to students worldwide and awarded to the strongest original research article or academic paper in science, technology, engineering, or mathematics. Entries were evaluated on the depth and originality of the research, the quality of the analysis, the clarity of the writing, and the potential significance or impact of the work.

The winning paper

Amshul's paper, Can a fruit-fly compass circuit make a cheap robot navigate better? A simulation study with a hardware validation protocol, starts from a clean engineering question. The fruit fly's central complex maintains a ring-shaped "compass" of neural activity that tracks heading, integrates turning, and is corrected by landmarks; a downstream circuit compares that heading with a goal direction to steer left or right. Does copying that architecture actually help a low-cost robot, or is it just biologically charming?

To find out, the paper builds the fly-inspired circuit as a 16-unit ring attractor and tests it against two baselines given the same sensors: a memoryless reactive controller and a conventional complementary-filter heading estimator. All three controllers share one obstacle-avoidance layer and were run through 4,200 simulated trials (100 random arenas, seven sensor-fault conditions, and two arena types) with paired statistics.

What it found

  • Against the conventional filter, the fly-inspired controller matched it on success (+0.0 percentage points in both open clutter and barrier arenas) and made slightly fewer collisions (0.09 and 0.17 fewer per trial).
  • Against the reactive controller, it raised success in open clutter by 1.4 percentage points and cut collisions from 1.71 to 0.32 per trial, with no significant success gain in barrier arenas.
  • In an isolated heading test, the ring was about 1° worse than the filter with sparse landmarks and 0.3° better with dense ones.

In the paper's own words:

"The biological architecture is a competitive but not superior heading system in this task, at an estimated 4–9× the compute of the filter."

It also found that the main failure mode under sensor loss "lay in the shared avoidance layer, not the heading circuit."

What stood out

The paper is candid about what it has and has not shown. Everything measured comes from a physics-based software simulation; the hardware experiment, a roughly US$55–65 Arduino robot, is written up as a validation protocol with predictions rather than presented as a result. And the takeaway is a careful one: the benchmark "shows that claims of biological advantage should be tested against a matched conventional baseline and that fault tolerance in low-cost robots may hinge on the avoidance layer."

Congratulations again to Amshul on the paper and on the scholarship.

Sources

  • Amshul Medisetti, "Can a fruit-fly compass circuit make a cheap robot navigate better? A simulation study with a hardware validation protocol" (2026), winning entry, Economic Mosaic STEM Research Scholarship
  • Green et al., "A neural circuit architecture for angular integration in Drosophila," Nature 546, 2017 — nature.com
  • Turner-Evans et al., "Angular velocity integration in a fly heading circuit," eLife 6, 2017 — elifesciences.org
  • Hulse et al., "A connectome of the Drosophila central complex reveals network motifs suitable for flexible navigation and context-dependent action selection," eLife 10, 2021 — elifesciences.org
  • Westeinde et al., "Transforming a head direction signal into a goal-oriented steering command," Nature 626, 2024 — nature.com