During an aerial survey flight, the UAS detected a rapid loss of battery power and initiated an immediate automatic landing. Whilst descending, its flight behaviour became erratic, control was lost to the remote pilot, and the UA struck an uninvolved person before hitting the ground. The photograph shows the UA after falling to the ground.

It was not possible to determine a cause for the loss of battery power or the flight behaviour. There may have been an opportunity during flight control checks to consider unexpected battery discharge rate as a reason to abort the flight.

The flight was planned in accordance with applicable regulation and guidance for safe distances between the UA and uninvolved people and structures. Due to the erratic flight behaviour of the UA during its automatic landing, coupled with loss of flight control, the UA subsequently breached the required safe distances and then struck the construction worker. The battery was fully charged prior to the flight, was within the manufacturer’s recommended number of cycles, and no defects were found to have affected its chargeDuring flight, the battery percentage detected by the UAS dropped at an unexpectedly high rate that resulted in an automatic landing without notifying the pilot at the pre-set charge thresholds. The first indication of an abnormal battery discharge rate was during flight control checks. The pilot chose to continue with the planned flight as it would have been achievable using the indicated 88% battery level. Subsequently, the UAS continued to detect a high discharge rate which triggered the automatic landing. After the flight the battery integral charge level indicator showed 50-75% charge. It was not possible to determine the difference between this charge level and the battery power displayed on the handheld controller during the flight. Flight log data did not explain the discrepancy in battery level.

The UAS measured a high loss of battery power in a short space of time, leading to an uncontrolled landing where the distance between the UA and uninvolved people and structures was compromised. This sequence of events resulted in the UA striking an uninvolved person. It was not possible to determine a cause of the UA’s detecting a loss of battery power or its flight behaviour. Whilst the displayed level of power remaining was sufficient for the planned flight, there may have been an opportunity during flight control checks for the pilot to consider the abnormal battery discharge rate was likely to continue, and to abort the flight.

The flight crew were contracted to provide drone footage for a TV drama production. They had both worked with the production company on previous series of the drama since 2017, and with the producer since 2021. On the day of the incident the flight location was an exposed headland with cliffs, with the task to provide “top down” views of the cliff scenery to be used for VXF (visual effects), then later on in the day to film an actor jumping onto a green screen. The flight crew (remote pilot and camera operator / observer) normally fly a DJI Inspire 2 UAS with this production, however the wind speed was above the Inspire 2’s rated wind resistance of 10m/s, so the Mavic 3 Cine was chosen for the task (it is also far more stable for top-down shots than the Inspire 2).

It is normal procedure for the camera operator to connect the remote controller (RC) to an external monitor for the production crew to view during the flight. The Mavic 3 Cine does not allow a separate camera operator, but the DJI RC Pro does have an HDMI port to output to an external monitor. The RP had used this setup before (the last time was November 2022), however on the day of the incident the HDMI connection did not appear to work. The DJI RC Pro has a mini HDMI socket and the flight crew only had a single cable of this type so could not try a different cable (this was the same cable used in November and had not been used since). The flight crew discussed the problem and thought it may be due to a firmware update. Searching the web there were other similar instances, and answers pointed to an HDMI setting on the device, however this did not work.

It had now taken almost half an hour trying to diagnose the problem and the producer was wanting the drone in the air as soon as possible because the other part of the shoot was ready and the actors on their way to the set. The RP explained the problem and the producer was happy to look over the RP’s shoulder at the RC Pro screen. The UAS was prepared for the task and the normal checks completed, however the RP failed to check the action on loss of C2 link. The UAS was recently flown from a boat so the loss of C2 action was set to hover instead of return to home (RTH).

With the Inspire 2, the production crew would be looking at the monitor and discussing the shots with the camera operator, who would also give instructions to the RP to position the UA. The RP would keep the UA in VLOS at all times, and report back to the camera operator if a particular manoeuvre was not possible or would impact VLOS. On this flight the RP was having to look at the screen and discuss the shot with the producer whilst the camera operator – now acting as a trained UA observer – kept VLOS with the UA reporting back to the RP. The producer requested the RP to lower the UA height at an ideal spot for the shot. As the UA descended it went behind a higher section of the cliff, and the camera operator immediately reported to the RP that he had lost VLOS. In the same instant the C2 link dropped, and the RP was unable to regain control the UA.

The RP reported that the UA would initiate RTH after a few seconds then he could resume control, however the UA did not appear to rise up above the cliff. It was at this point the RP realised he had not checked the C2-loss action, and it was likely set to hover from the work off a boat. The RP and observer started walking towards the UA location, although the terrain was very steep and difficult. By the time the RP was about 50m from the UA’s location the low battery failsafe RTH commenced and the C2 link restored as soon as the UA ascended above the cliff. He landed the UA and returned to the production area, there was no damage to the UA and on landing the battery was at 17% remaining charge.

Given the remote location and proximity to the cliffs there was no danger to uninvolved people or other aircraft during the time the UA was hovering. The remainder of the day’s flights were executed without any issues.

Lessons learned: The RP was under pressure and failed to check the loss of C2 action, even though it is listed as a check item in the FRCs. The company daily risk assessment document will be updated to ensure all checks are completed prior to flight.

Due to low temperature (-3c) battery simply stopped operating and the drone fell. In subsequent correspondence, the reporter added that the drone was a DJI phantom 4 pro. A great bit of kit, well supported by DJI with software and updates. Yes, indeed it was -3, with several inches of snow.

When I couldn’t find the drone after it disappeared, I hooked my base unit up to my PC at home, and yes, it had recorded the entire flight, and I was able to locate it in daylight hours, the following day. I sent it off for repair and paid about around £200 for gimbal damage.

The drone was 2 years old when it had the event in the snow. Hours-wise and battery-wise, it had flown about 40 hours. From memory I think the batteries had been recharged around 30 times. The aircraft had all auto functions switched on and despite having flown quite a few trips over some of our local lochs, there was never any indication of issues caused by faulty inputs.

Sadly, I no longer have the drone as I experienced a detachment of a rotor blade during a subsequent flight, and although I located it using the DJI app, it was in such a heavily forested area, I lost it altogether. I have more than 15,000 hours flying all sorts of aircraft in all sorts of situations, military ops, difficult terrain etc, and I am more careful than needed. I also teach human factors, so feel I could recognise such, should they appear. I don’t think they affected the operation.

Bottom line, imho, (in my honest opinion) low oat, battery performance, sudden death of the battery. Don’t fly in cold weather. Software located drone (it was white against a background of 1 foot of snow)

Following their GVC (General Visual Line of Sight Certificate) training, a newly qualified Remote Pilot (RP) experienced a minor accident whilst conducting the 6 hours of non-operational mandatory flight training that precedes the [XXXXX] flight assessment. On Sunday 7th May, the pilot was flying a DJI M30T in a field >50m away from a railway. The flight was taking place at night, which appears to have been a contributing factor in the accident.

The RP was practicing a rapid descent deconfliction manoeuvre using a DJI M30T, and had rapidly descended the aircraft from an altitude of approximately 80m to 3m. The RP simultaneously reduced the lateral distance of the aircraft from approximately 390m to 3m. At the end of the descent manoeuvre, the RP accidentally pushed forward on the right control stick, instead of the left (throttle) and the aircraft collided with a nearby palisade fence. The aircraft experienced minor damage, albeit adequate to necessitate repair by an external service provider.

Whilst it is generally considered positive that the RP was practicing an avoiding manoeuvre in a rural location, the RP descended the drone at excessive speed, leading to confusion regarding the control stick inputs. The RP pushed his flying skills too far, too fast during training and should have avoided such actions during a night flight.

The RP was undertaking a training flight at night – for this reason, the lighting conditions were insufficient for the obstacle avoidance sensors to function and, as a failsafe, avert the collision. The RP was also alone during the training flight – whilst this is permitted, it is advised that training flights should be conducted with an experienced colleague where possible and available for support.

Operator’s internal recommendations:

• • Only conduct simple flights at night and be aware that obstacle avoidance sensors will not work unless there is sufficient lighting.
  • You are obligated to read your aircraft manuals and must understand how the aircraft will behave under different weather and lighting conditions.
  • Night Operations require robust planning, risk assessments and mitigations to be in place.
  • These include, but are not limited to, the use of TOAL lighting and minimising the flight distance. You are also required to observe, and risk assess the site during the day.
  • It is important not to push your skills too far, too fast – particularly if you are a newly trained RP. It is advised that you should identify and contact a more experienced colleague to support you during training flights.

We are grateful to the BMFA that they have now established a system whereby we automatically receive incident reports as they are filed. The occurrence happened above 400ft (120m) and the unmanned aircraft flew beyond visual line of sight and was not found.

The Model was an EFlite Mystique RES glider, flying weight ~5lb, wingspan 2.9m. Conditions were clear blue sky with small patchy clouds, wind from NE around 8mph gusting 15mph? I had climbed to ~500 ft and was circling in what I believed was lift centred over the flying patch. My telemetry was reporting current height just over 700 ft when the glider suddenly became indistinct and then totally disappeared against what I had taken to be a clear blue sky. After a few seconds of panic, I activated the spoilers and held in rudder, but the model did not reappear.

The next height report (30 seconds after the 700+ ft) was 79ft but I did not see or hear anything impact nearby and was concerned that the model actually carried on climbing and that the 79ft was a false reading. If the model had lost 700ft in 30 seconds, I would have expected to hear wing flutter on this model and then the noise of an impact. I did not see the model again and the telemetry lost connection. About 30 minutes later on, I drove downwind along the road for ½ mile or so, but the model did not reconnect with the Tx, so I have no idea where it may have ended up.