Learnings
Outcomes
WP1 Design Planning & Reporting
● NG visit to Lavant. Sees shared system capabilities, NG performed VLOS flight around pylon for use case discussion
● Project requirements baselined and 1st milestone report submitted
● Annual NG NIA progress report completed for NG submission
WP2 sees.ai Solution Development
● Comms developments for improved robustness and longer range flights
● Software development improving both backend and front end useability
● Hardware development with the introduction of a new UAS, improving robustness and flight time
WP3 Integration into NGET workflow & tools
● Use case development working with NG
● Sees software mission enhancements
● Updated mission template specification based on specific NG assets
● Further template improvements during the project reducing time to set/change parameters for a new pylon and improving angles of image capture for better results
WP4 NGET Testing and Evaluation
● 3 trials executed over the course of the project providing the necessary levels of quality and coverage. These trials were conducted at two different sites, Keadby and Chalton.
○ Tension and suspension towers for fittings and steelwork remained the core focus of the project. Terminal towers were also included, flying manually.
○ The team focussed on capturing detailed images of end fittings, jumper cable asset / spacer inspections, ‘dolly sets’: extra insulators on corner towers all at the correct angles for maximum coverage.
○ Two approaches were explored further 1) using a longer lens to get even more detail and tighter framing and 2) flying faster using linear scans to help increase the run rate.
○ During the trials, the team developed a river crossing steelworks template, and completed a span inspection feasibility test.
○ During trial 2, 34 towers were inspected in 2 days. 22 of these were inspected in one day.
○ Run rate exercises were performed throughout the trials to measure the number of towers that could be surveyed in a business day. Improvements were made between trials which dramatically improved run rate stats.
■ In summary:
● Sees were able to survey two towers per flight from trial 2 thanks to architectural and functional changes that were made to the automated missions system.
● Survey rate increased from 9.4 to 16.5 towers per day
● Average time per tower reduced from 24 mins to 13.5 mins
● Further details can be found under Key Learnings below.
WP5 Advancing Aviation Permission:
● Sees were successful in gaining further BVLOS approval from the CAA for non geo-specific BVLOS operations with visual observer mitigation (no local pilot needed).
● Sees submitted a variation to authorise the use of their recently developed airframes (YEALM)
● Sees submitted a variation that introduces the concept of a Transit from the TOLZ to the Atypical Air Environment flight corridor. This was a new concept for the CAA and it addresses the limitation of having to place the TOLZ within the AAE which is not always practical or possible.
● Sees will be sharing insights and learning from the OSC approval process with NG in due course to assist with any permissions they will be seeking in the future.
WP6 Extend to Long Distance Flight
● Using the software enhancements and developments during the project Sees were successful in stretching the existing system to fly longer distances by improving comms range, flight time and operations efficiency.
This culminated in the long distance test surveying 2km of the 4VF line. It was an automated mission capturing 6 images at each of the 7 towers and it took just over 24 mins.
WP7 Data Transfer to third party data suppliers
● Data (video, payload imagery and metadata) was transferred to the processing platform (Keen.ai) using NG BAU transfer process using GCloud for onward processing, assessment and evaluation.
Lessons Learnt
The following are the Key Learnings from this project related to sees.ai’s work (i.e. related to data capture and delivery):
General:
● Run Rate:
During the 3 trials, a key area to understand was the number of pylons that could be successfully surveyed by the ops team in a day, to ensure that challenges were exposed and understood and could be addressed or improved upon during the project. Full details can be found in the 2 run rate reports - see below.
○ Run Rate 1 (See Appendex A2)
● Survey rate = 9.4 towers per day
● Average time per tower = 24 mins (Suspension)
○ Run Rate 2 (See Appendex A2)
● Survey Rate = 16.5 towers per day
● Average time per tower = 13.5 mins (Suspension)
The following areas of improvement were identified and worked on (or are planned to be) over the course of the run rate exercises and will be further enhanced going forward:
○ Battery swapping, docking and charging
○ Airframe weight
○ Further template improvements
○ Gimbal / drone and camera control updates reduce time per image.
○ Automated image processing to improve post flight admin time.
○ Further increase in comms range, beyond 800m
● Reliance on Wayleave / land access
Access to land/wayleave in order to survey pylons was managed by NG and their external consultants. Only a subset of permissions were received in time for the final trial which meant that not all of the pylons could be surveyed. The project team are working with NG to look at opportunities to improve the process going forward.
Out of the 61 towers 4VF29-89 for which access was requested for the 3rd and final trial, 38 were granted of which 14 could only be accessed for two consecutive days due to 3rd party game shooting arrangements.
The location and land access in Trial 3 highlighted key limitations:
○ The reliance on finding suitable takeoff sites on public land which could be used with Sees current comms system (substantial LoS required)
○ The requirement for a VLOS safety pilot as the current Sees drones are not yet compliant with the BVLOS provisions as per their Ops manual. Discussions are underway to collectively review the current process and look for further efficiencies. Drone enhancements underway to meet BVLOS compliance (focussing on Comms)
● Business case:
○ During the requirements gathering process and then throughout the trials, valuable insights were gained from NG which steered the course of the project to ensure the outcomes were directly addressing the business need and BAU challenges. Focus was placed on capturing the right components at the best angles, to drive for accuracy and efficiency over speed at all costs. With improved accuracy and efficiencies, survey rate naturally improved.
Technical:
● Mission development:
○ Mission adjustment can be very time consuming especially if using the current s/w parameter sliders: the position on the screen means that the possible adjustment range for click and drag is very asymmetric.
● Mission backend/front end enhancements:
○ Mission can’t be edited if the drone breaches a loaded georestriction.
○ Time required to generate new missions / build new templates was too long.
○ Execution of the task sequence for image capture was too long for BAU operations.
As a result of the above, Improved mission pre-planning and template generation was developed and implemented during the project.
Parallel task sequencing implemented during the project.
● Long range flights (learning and subsequent development during the project):
○ Improved drone sensor calibration - resulting improved 3D world accuracy.
○ Improved gimbal tuning - resulting in better image framing.
○ Various mission system bug-fixes, resulting in a more stable and reliable mission system.
○ Longer range flights awaiting improved comms system.
● Image quality and camera gimbals:
○ Images were over-exposed in the early flight tests in the project and on the rear cameras, image sharpness was low compared to the front.
○ Gimbal targeting was problematic at times, impacting the image framing - meaning that some targets could not be captured.
As a result of the above, framing, camera and gimbal adjustments were made which resulted in improved imagery across the board.
● Communications
○ Throughout the majority of the trials, intermittent comms issues were experienced which led to the Sees team refining their set-up both fronthaul, between the drone and the mobile GCS van, and backhaul from the mobile GCS back to HQ.
