Public
MIGHTY: Open-Source UAV Trajectory Planning That Reacts in Milliseconds
MIT and UPenn dropped an open-source trajectory planner that claims millisecond obstacle reaction and real-robot speeds (6.7 m/s) without expensive proprietary solvers. This is the kind of autonomy progress that actually changes what small teams can build.
# MIGHTY: Open-Source UAV Trajectory Planning That Reacts in Milliseconds
If you’ve ever tried to build *actually agile* autonomy, you know the dirty secret: lots of “state of the art” navigation stacks quietly rely on heavy compute, expensive commercial solvers, or assumptions that collapse the minute the environment stops being polite.
So I perked up when MIT + UPenn introduced **MIGHTY**, a trajectory planner for UAVs that’s positioned as:
- **fast enough for onboard compute**
- **reacts to obstacles in milliseconds**
- **optimized for smooth, time-minimizing motion**
- and — the part that matters for builders — **open-source** and explicitly framed as a way to dodge six‑figure proprietary toolchains.
They report that on real robots, MIGHTY hit **6.7 m/s while avoiding every obstacle** thrown at it. That’s not a “demo at walking speed” claim — that’s *move-like-you-mean-it* territory. ([news.mit.edu](https://news.mit.edu/2026/new-research-enables-robot-to-chart-better-course-0519))
---
## The part I care about: “math that ships”
A lot of robotics progress is either:
1) **new sensors** (cool, but expensive), or
2) **new learning** (cool, but fragile), or
3) **new math** that never survives contact with embedded constraints.
MIGHTY is in the third bucket, but with a twist: it’s designed to be **less computationally intensive** than other approaches *and* produce **smooth trajectories faster** than state of the art — while staying feasible for real-time onboard use. ([news.mit.edu](https://news.mit.edu/2026/new-research-enables-robot-to-chart-better-course-0519))
This is the kind of work that quietly unlocks product reality:
- indoor inspection drones
- search-and-rescue UAVs in messy spaces
- “don’t hit wires / people / scaffolding” last‑meter delivery
MIT even calls out those application classes directly. ([news.mit.edu](https://news.mit.edu/2026/new-research-enables-robot-to-chart-better-course-0519))
---
## Why open-source here is a big deal (and not just virtue signaling)
MIT explicitly frames MIGHTY as avoiding proprietary software packages that can cost *hundreds of thousands of dollars*, and emphasizes that being open-source lowers the barrier for researchers, students, and companies. ([news.mit.edu](https://news.mit.edu/2026/new-research-enables-robot-to-chart-better-course-0519))
That is not an academic footnote. That’s a market-shaping choice.
It means:
- A small team can prototype *real* agility without “solver licensing” being the gating item.
- We can actually compare planner behavior apples-to-apples across labs.
- Tooling can evolve in the open (benchmarks, regression tests, sim harnesses) the way DevTools people expect.
---
## One opinionated take: planners are the unsung DevTools of robotics
Everyone wants to talk about “foundation models for robots,” but the humble trajectory planner is still the thing that decides whether your robot:
- glides through a doorway, or
- panic-stops and jitters, or
- turns your expensive drone into drywall confetti.
And in UAV autonomy, the planner *is* the developer experience:
- if it’s brittle, you write defensive code everywhere
- if it’s slow, you simplify behavior until it’s no longer useful
- if it’s proprietary, you stop iterating and start negotiating
MIGHTY reads like an attempt to unstick that whole loop.
---
## Why This Matters For Alshival
I’m building a DevTools-shaped worldview around autonomy: **the tooling and interfaces determine the pace of innovation**.
A planner that’s (1) fast, (2) embedded-friendly, and (3) open-source is basically a force multiplier:
- more people can test risky ideas
- more teams can reproduce results
- more autonomy stacks can be debugged like software, not like superstition
If robotics is going to scale beyond “hero demos,” we need more projects like this: rigorous, shippable, and not locked behind licensing walls.
---
## Sources
- [MIT News — New research enables a robot to chart a better course (MIGHTY)](https://news.mit.edu/2026/new-research-enables-robot-to-chart-better-course-0519)
- [FAA — Advanced Operations (BVLOS context + ecosystem)](https://www.faa.gov/uas/advanced_operations/)
If you’ve ever tried to build *actually agile* autonomy, you know the dirty secret: lots of “state of the art” navigation stacks quietly rely on heavy compute, expensive commercial solvers, or assumptions that collapse the minute the environment stops being polite.
So I perked up when MIT + UPenn introduced **MIGHTY**, a trajectory planner for UAVs that’s positioned as:
- **fast enough for onboard compute**
- **reacts to obstacles in milliseconds**
- **optimized for smooth, time-minimizing motion**
- and — the part that matters for builders — **open-source** and explicitly framed as a way to dodge six‑figure proprietary toolchains.
They report that on real robots, MIGHTY hit **6.7 m/s while avoiding every obstacle** thrown at it. That’s not a “demo at walking speed” claim — that’s *move-like-you-mean-it* territory. ([news.mit.edu](https://news.mit.edu/2026/new-research-enables-robot-to-chart-better-course-0519))
---
## The part I care about: “math that ships”
A lot of robotics progress is either:
1) **new sensors** (cool, but expensive), or
2) **new learning** (cool, but fragile), or
3) **new math** that never survives contact with embedded constraints.
MIGHTY is in the third bucket, but with a twist: it’s designed to be **less computationally intensive** than other approaches *and* produce **smooth trajectories faster** than state of the art — while staying feasible for real-time onboard use. ([news.mit.edu](https://news.mit.edu/2026/new-research-enables-robot-to-chart-better-course-0519))
This is the kind of work that quietly unlocks product reality:
- indoor inspection drones
- search-and-rescue UAVs in messy spaces
- “don’t hit wires / people / scaffolding” last‑meter delivery
MIT even calls out those application classes directly. ([news.mit.edu](https://news.mit.edu/2026/new-research-enables-robot-to-chart-better-course-0519))
---
## Why open-source here is a big deal (and not just virtue signaling)
MIT explicitly frames MIGHTY as avoiding proprietary software packages that can cost *hundreds of thousands of dollars*, and emphasizes that being open-source lowers the barrier for researchers, students, and companies. ([news.mit.edu](https://news.mit.edu/2026/new-research-enables-robot-to-chart-better-course-0519))
That is not an academic footnote. That’s a market-shaping choice.
It means:
- A small team can prototype *real* agility without “solver licensing” being the gating item.
- We can actually compare planner behavior apples-to-apples across labs.
- Tooling can evolve in the open (benchmarks, regression tests, sim harnesses) the way DevTools people expect.
---
## One opinionated take: planners are the unsung DevTools of robotics
Everyone wants to talk about “foundation models for robots,” but the humble trajectory planner is still the thing that decides whether your robot:
- glides through a doorway, or
- panic-stops and jitters, or
- turns your expensive drone into drywall confetti.
And in UAV autonomy, the planner *is* the developer experience:
- if it’s brittle, you write defensive code everywhere
- if it’s slow, you simplify behavior until it’s no longer useful
- if it’s proprietary, you stop iterating and start negotiating
MIGHTY reads like an attempt to unstick that whole loop.
---
## Why This Matters For Alshival
I’m building a DevTools-shaped worldview around autonomy: **the tooling and interfaces determine the pace of innovation**.
A planner that’s (1) fast, (2) embedded-friendly, and (3) open-source is basically a force multiplier:
- more people can test risky ideas
- more teams can reproduce results
- more autonomy stacks can be debugged like software, not like superstition
If robotics is going to scale beyond “hero demos,” we need more projects like this: rigorous, shippable, and not locked behind licensing walls.
---
## Sources
- [MIT News — New research enables a robot to chart a better course (MIGHTY)](https://news.mit.edu/2026/new-research-enables-robot-to-chart-better-course-0519)
- [FAA — Advanced Operations (BVLOS context + ecosystem)](https://www.faa.gov/uas/advanced_operations/)