Akara Robotics

Nominated Award: Best Application of AI to achieve Social Good
Website of Company: https://www.akara.ai/
Akara is a robotics research and deployment company. Our mission is to develop automation to empower frontline healthcare workers to make hospitals safer and more efficient. Through better use of robotics and AI technologies, we’ve shown that we can help hospitals substantially reduce downtime in critical areas (including radiology, endoscopy, operating rooms, among others) while enhancing environmental cleanliness, reducing operational costs, and reducing the occupational risk faced by frontline healthcare workers. Our technology has been validated through numerous peer-reviewed scientific publications and is currently in active use by the Irish Health Service (HSE) and the NHS. Akara is a spin-out from one of Europe’s leading universities. Our award-winning technology has previously featured on the cover of Time magazine.
Akara is a spin-out from Trinity College. The Akara founders have been working together for more than 7 years and collectively have more than 40 years experience developing service robots. Akara co-founder and CEO, Conor McGinn, was formerly listed by MIT Technology Review as one of Europe’s top 35 innovators under 35. Niamh Donnelly, co-founder and chief robotics officer, was recognized globally as one of “50 women in robotics you need to know about 2021” by Robohub and was recently awarded the Woman in Technology prize at the Women Mean Business awards.
Reason for Nomination
Problem
The COVID-19 pandemic has highlighted two major weaknesses in hospital infrastructure. Firstly, it revealed how poorly equipped hospitals are to stop the spread of pathogens in the air. This was evident by the high incidence of COVID-19 infections among healthcare workers; at points during the pandemic, one in three confirmed cases in Ireland were healthcare workers. And secondly, it revealed how inefficient hospital decontamination protocols are. We saw this through the impact that additional room cleaning requirements had on hospital workflow. For example, arising from a need to decontaminate treatment rooms on a frequent basis, radiology departments reported a 50% reduction in scans during periods of the pandemic.
At the core of the problem is that hospital disinfection remains dependent on manual cleaning methods that are known to be time consuming, inefficient, and hazardous to the health of environmental services workers. For illustration, the time to disinfect a radiology treatment room was reported to be in the region of 30-60 minutes during the pandemic. Considering that a typical radiology scan takes 10 minutes to perform, this had a dramatic effect on the number of patients that could be treated each day. Factors contributing to this time included waiting for cleaners to arrive, the time to prepare the cleaning solution, the time to manually administer the solution, and then the time to allow the cleaning agent to dry, which is normally 10 minutes. Manual disinfection has known reliability issues. Since germs are too small to see with the naked eye, the cleaner cannot easily tell the parts of the room where contamination levels are greatest, and they simply do not have the time or ability to disinfect everywhere to the same extent. Also, the efficacy of the chemical agent is affected by several factors including the surface material, the concentration or wetness of the chemical at the surface, and the genetic structure of the microorganism. Studies conducted on the efficacy of manual cleaning suggest that it may only be effective at removing just over 50% of the target pathogens.
Cleaning chemicals are toxic and they are known to cause harm if exposed directly to skin or eyes. When mixed with water, as is the standard practice in hospitals, they also produce a poisonous gas which is linked with respiratory illnesses. Research has found that environmental services staff are disproportionally more likely to develop asthma and other respiratory illnesses due to this occupational exposure. Furthermore, these chemicals are known to be especially harmful to the environment, especially when they enter waterways through the sewerage system. Current Technologies Devices that use ultraviolet light have emerged as a potential solution to this problem. At certain wavelengths in the ultraviolet spectrum, light is known to have germ killing properties. However, since these devices required must be pushed into place and can only be used in rooms that were empty (because direct exposure to UV light can be dangerous), there use necessitates additional staffing and they lack the ability to reduce room downtime.
Our Innovation
Akara are developers of Violet, the first clinical-grade fully autonomous UV disinfection robot. Violet is a mobile robot comprised with advanced sensing and AI abilities that allow it to traverse complex hospital environments. It has three powerful UV lamps mounted in a rotating column, enabling it to closely control the parts of the room that are irradiated. A video of Violet can be found at the following link https://youtu.be/IvJqWwDxxkk
Violet has two core unique selling points, both enabled by AI, that differentiate it from the competition. First, whereas competitor technology requires the robot to be manually pushed into place in the room, Violet can move autonomously. During a room decontamination procedure Violet navigates to up to 30 locations in the room. This ensures greater room coverage (since the likelihood of shadowing is reduced) and means that the disinfection procedure can be carried out with virtually zero additional staffing requirements. And second, Violet is the only disinfection robot that doesn’t require the room to be evacuated during use. Since staff can be present at the same time as the robot, they can prepare the room for the next patient while the robot operates. This leads to a very significant reduction in room downtime. This functionality is enabled by a computer vision algorithm running on an edge device that
constantly tracks the location of people in the room during the disinfection procedure. Since the robot can control which parts of the room are irradiated in real-time, if a person gets too close to the irradiation zone, the robot automatically redirects or turns off the lights.
Validation
In January 2020, we became the first to validate the efficacy and feasibility of a UV disinfection robot in peer-reviewed scientific journal (available at https://www.frontiersin.org/articles/10.3389/frobt.2020.590306/full ). This paper has since amassed more than 15,000 reads placing it in the top 5% of papers in the Frontiers journal catalogue. Included as supplementary material to this paper is a video showing the autonomous navigation system in action, as well as a dynamic path planning whereby on detection of an obstacle in its path it was able to generate a new trajectory to its goal.
In January 2022, we published a paper in the American Journal of Infection Control (top journal in the field according to Google Scholar) demonstrating the ability for our UV disinfection robot to achieve improved room decontamination performance in less than 25% of the time that it normally takes to perform a manual room clean. This involved the development of significantly improved robot platform and major updates to the underlying navigation algorithm. A video of the robot in action can be seen at https://youtu.be/GVNzdstppNE
Societal impact
Our technology achieves social good because, its use • lessens cleaners exposure to harmful chemicals • reduces the amount of toxic chemical waste that ends up in waterways • timesavings from one robot can enable up to 6000 additional procedures a year in radiology and up to 2500 additional procedures a year in endoscopy
Additional Information:
Funding Milestones
In March 2021, Akara closed a 250k pre-seed funding round. Investors included a well-known super angel. In October 2021, Akara were awarded 2.5M equity- free funding by the European Innovation Council (EIC). The EIC program is Europe’s most competitive early stage deep-tech funding program.
Technical Summary of AI on Robot
Our robotic software programs run within a Robot Operating System (https://www.ros.org/) environment and are written in a combination of C++ and python. Our robot generates 2D occupancy grid maps from a combination of sensors including a 2D lidar (Hokoyu URG) and four Intel RealSense RGBD cameras. Our autonomous navigation algorithms can be configured to use a range of global planners, including the A* algorithm.
Our robots have four cameras which give it a 360 degree view of the robot’s surroundings. Each camera is connected to an Edge device (Luxonis) thus preserving computational bandwidth of the main on-board PC. Person detection is implemented in two way, either (1) using a pre-trained object classifier model that we have modified to work on the edge device, or (2) using a custom trained CNN that we have created. We have found that the former approach generally achieves the best overall performance, however the custom trained model is needed in settings where staff wear hazmat suits and aren’t easily detected by the standard classifier.