Oakland Taps DJI Mavic 3 Drones to Fight Illegal Dumping

Oakland is about to become the second U.S. city to run a DJI Mavic 3 Enterprise on AI-powered garbage patrol. The city council is moving forward with a six-month, $150,000 pilot contract with San Francisco startup Aerbits, which will fly 72 missions over roughly 1,440 miles of Oakland streets, as reported by NBC Bay Area.

The system uses computer vision to spot mattresses, tires, furniture, and construction debris, then files 311 service requests automatically with GPS coordinates and timestamped aerial photos attached.

This is not a surveillance program. It’s a logistics tool with receipts.

The Hardware Doing the Work

Aerbits officially confirms the DJI Mavic 3 Enterprise as the airframe behind the entire operation. That hardware choice is not a coincidence. The M3E was purpose-built for exactly this kind of systematic mapping mission.

The specs matter here. The Mavic 3 Enterprise carries a 20-megapixel Four Thirds CMOS wide camera with a mechanical shutter, which eliminates motion blur on mapping flights, and it triggers a photo every 0.7 seconds.

Flight time hits 45 minutes maximum, top speed runs about 33 mph, and with the optional RTK module the aircraft achieves centimeter-level positional accuracy. A 56x hybrid zoom gives the AI model enough pixels on target to classify waste types reliably.

DJI rates the airframe for covering up to 0.77 square miles per mapping flight, which lines up exactly with what Aerbits does on the ground: one square mile of coverage every 30 minutes.

That’s the efficiency difference that makes this model work. Traditional ground-based enforcement covers about two city blocks per hour. The M3E covers a full square mile in the same window.

The Bayview Data Is Stronger Than Press Release Math

Aerbits is not showing up to Oakland with slides and promises. The company ran a 13-month self-funded pilot in San Francisco’s Bayview-Hunters Point neighborhood from March 2022 through April 2023. The numbers are real and the methodology is more rigorous than anything else in the illegal dumping technology space.

Across the full program, Aerbits flew 125 missions, captured 117,793 aerial photos, identified 4,441 unique dumpsites, and filed 4,376 service requests to SF 311. Peak reduction hit 96 percent.

The more interesting piece of the Bayview data is a 52-day stretch in May and June 2022 that Aerbits structured as an A-B-A withdrawal study. That’s a recognized quasi-experimental design where a treatment gets applied, pulled, then reapplied to test whether the observed effect actually traces back to the intervention.

The baseline on day one counted 118 active dumpsites. After 26 days of daily flights, that dropped to 85. Then Aerbits paused monitoring for 14 days. Active dumpsites rebounded to 91, almost back to baseline. When flights resumed for another 24 days, the count fell to 5.

The rebound is the whole story. If the initial drop had been caused by anything other than the monitoring, warmer weather or a random city cleanup push or seasonal variation, dumpsites would not have surged back the moment the drone stopped flying. That tight coupling between flight status and dumpsite count is what turns the Bayview numbers from a testimonial into evidence.

The Reports Residents Never Filed

One of the more uncomfortable findings from Bayview is the detection gap. Aerbits reports that 30 to 50 percent of the dumpsites the drone found on public roads and sidewalks had never been reported through 311 at all.

Not because residents didn’t care. Because some sites aren’t visible from commonly traveled paths, because residents face language barriers, because people have learned that complaints don’t lead to action, and because complaint-driven systems structurally favor the neighborhoods with the highest trust in city government.

District 2 Councilmember Charlene Wang made this case directly at the Oakland committee meeting. She said the San Antonio and Little Saigon neighborhoods of her district, where she describes the dumping as the worst, generate only about five 311 reports in a given window. Wealthier parts of her same district file around 400 reports for far smaller problems.

Aerial coverage flips that math. The drone looks at every street the same way regardless of who lives there. That’s the environmental justice argument, and it’s the strongest non-efficiency reason for cities to take this seriously.

Oakland’s Privacy Guardrails

Councilmember Zac Unger framed the program carefully. The drones identify dumping hotspots and help Public Works dispatch the right truck size. They are not capturing faces or license plates, and the data is not being stored for enforcement purposes. Oakland’s Privacy Advisory Commission reviewed the proposal and approved it in March 2026 before it moved to the full council.

Oakland is separately considering an expansion of license plate readers to target dumpers directly, and Wang pointed out that the city has only 36 license plate cameras citywide.

That’s a different program with much heavier privacy implications. The Aerbits drones and the license plate readers should not get lumped together. One is a pattern-detection tool pointed at trash piles. The other is an identification tool pointed at people. Different mission, different oversight, different scrutiny.

DroneXL’s Take

Here’s what I find genuinely significant about this one. The Bayview A-B-A data is the first credible causal evidence I’ve seen in the illegal dumping technology space, and it was produced by one founder flying a Mavic 3 Enterprise on his own dime in a neighborhood that every other vendor ignored. That’s a better product development story than most venture-backed drone startups can tell.

The DJI angle is worth saying out loud because nobody at the Oakland council meeting mentioned it. This pilot runs on Chinese hardware, and that hardware is doing work that U.S.-built alternatives either can’t match at the price point or can’t match at all. The Mavic 3 Enterprise is the best tool for this mission because its mapping performance, its 45-minute endurance, its mechanical shutter, and its RTK accuracy are still the benchmark for compact enterprise drones.

Skydio doesn’t have an equivalent mapping platform in the M3E’s class. The American airframes that do exist cost substantially more per flight hour. If Aerbits had to respec this contract on non-Chinese hardware, the $150,000 price tag would look very different, and the 96 percent reduction figure might not be repeatable.

The honest sustainability question is whether Oakland’s pilot becomes a real program or a nice six-month demo. The Bayview data answers that question directly. When monitoring stopped for 14 days, the dumpsites came back. That’s not a minor caveat. That’s the entire operational model. Aerial detection works as long as it keeps flying. The moment the budget runs out, so does the cleanup.

Oakland should plan for that now, not in month six.

Photo credit: Aerbits