Meet Spot - Using drone technology to inspect assets
We are developing the use of robotics and drone technology to improve the quality of raw asset data collection and improve safety in the field.
We are using robotics and reinforcement learning (a form of artificial intelligence) to take our asset inspection to a new level this year. Spot, a robotic dog, represents the pinnacle of converging technologies. We are adapting the drone to suit our organisation by using customised software and recording equipment.
This is an investment to improve safety in the field and to develop skills within our organisation for the use of future drone technologies.
Spot uses ground-breaking technology like autonomous navigation, computer vision and smart robotics to navigate between our assets. We have placed an additional camera on board, that is used to inspect our assets.
Spot will move from Stobie pole to Stobie pole and take multiple images of the poles, powerlines and transformers as it moves along a street. Common issues to identify may include low lines, damaged poles, and ageing equipment. Spot will be able to navigate through rough terrain and areas that may be dangerous for staff. We will pair it with a skilled operator who will always stay near the unit and our staff will check the photographs to identify any issues.
By pairing a robot or drone with a skilled asset management officer, the officer can apply their skills where they’re best utilised. The robot meanwhile completes the repetitive, basic tasks like collecting images.
How does it work?
We’re currently training Spot. We’re using the programmable application software development kit (SDK) to teach Spot how to complete the tasks we need it to do. This is our first piece of artificial intelligence (AI) driven robotics software that we’re developing in-house. We’re using a relatively new technique called reinforcement learning. It’s a form of artificial intelligence (AI).
Reinforcement learning for Spot is much like teaching a pet. In the same way as you give a pet treats to teach it a trick, Spot is rewarded with numbers when it does what we want. Spot’s training is like playing a video game. If it does a good thing, such as take a photo of a pole, it gets a positive reward, and if it does a bad thing such as go onto the road, it gets a negative reward. It is always trying to get a high score.
We will provide Spot with geographic information such as latitude and longitude and Spot will be able to work out where to navigate. It can see the poles, navigate towards, and take pictures of them. Spot will wear a high-resolution camera that can pan, tilt and zoom, to photograph the assets and surrounding environment. A skilled asset management officer will then inspect the images. The officer will raise any maintenance requests as needed.
Current regulations do not allow the use of flying drones beyond the pilot's line of sight for asset inspections, however, eventually, this may be allowed. We need to prepare. A land-based drone is our best opportunity to do that. We can develop in-house capabilities, upskill our staff and get familiar with the technology now.
Spot has a range of features that suit our needs:
- Crash protection.
- Dynamic reaction.
- Self-rights after a fall.
- Average run time of 90 minutes with replaceable battery.
- 360-degree obstacle avoidance.
- Operational in temperatures ranging from -20 C to 45 C.
- Rain and dust protection.
SA Power Networks Leading Innovation
Our digital transformation plays a vital role in enabling SA Power Networks to deliver its strategic objectives and meet the challenges ahead. Using cutting-edge technologies like Spot, which combines robotics and AI, is one of the ways we’re striving to deliver better service to our customers.
We are proud to be the first utility in Australia to use Spot for meeting our strategic goals. Our investment in developing the technology to suit our needs will not only improve safety and efficiency across our network but is a strategic investment in emerging processes and technologies.