Energy efficiency in robotics software

Mining Energy-Related Practices in Robotics Software

“…The high discard-rate in this step is not surprising, but is rather in accordance with existing research confirming that developers tend to have limited knowledge of energy efficiency [37]. Mobile Issues Swanborn and Malavolta [50] Robotics Literature Cruz and Abreu [16] Mobile Commits, Issues, PRs Moghaddam, Lago, and Ban [35] Generic Literature Matalonga et al [33] Mobile System Events Chowdhury and Hindle [13] Generic Commit Messages Bavota [9] Generic Commit Messages Moura et al [36] Generic Commits Malik, Zhao and Godfrey [32] Generic Stack Overflow Procaccianti, Lago and Bevini [44] Cloud Literature Pinto, Castor and Liu [42] Generic Stack Overflow Procaccianti, Bevini and Lago [43] Cloud Literature Calero, Bertoa and Moraga [10] Generic Literature Pathak, Hu and Zhang [38] Mobile Issues, Forum Posts Phase 2 -In this phase we manually analyze all the 5,111 data points for removing false positives. This phase is performed systematically and iteratively by three researchers.…”

“…Other notable themes (40) -Other notable causes of energyrelated issues include: streaming images/videos (10 occurrences), interaction with sensors and actuators (9 occurrences), providing feedback to a GUI (8 occurrences), raw computation (7 occurrences), and networking (3 occurrences). Overall, the emerged themes are expected and most of them have been already identified as sources of energy-related issues in the literature, both for robotics software [50] and other domains like mobile apps [41], [42], [24]. This result of our study confirms that roboticists are facing in practice similar causes as in other application domains.…”

“…Robot navigation (11) -As shown in Section IV-A, mobile robots like rovers and drones are particularly discussed in the ROS ecosystem. As confirmed by the literature [50], one of the main energy consumers in mobile robots is robot navigation (e.g., robot acceleration, deceleration, turns). This phenomenon can be traced back to the fact that mobile robots heavily interact with hardware components that are particularly heavy on the battery, such as servos, motors, etc.. For example, in DP384 the well-known ros-planning/navigation2 ROS package is extended with a new feature for configuring the maximum speed of the robot at runtime is proposed as a way to reduce the energy consumption of the robot.…”

“…External validity -Despite the key role played by ROS in robotics software [46], [50], its coverage of a large variety of robots [23], and its vibrant open-source community, we are aware that ROS-based projects may not cover all types of robots. Our population data, however, is diverse (∼340k data points, including 10 different robot families and 12 distinct robot capabilities).…”

“…In this context, even a slight energy improvement can lead to great benefits in terms of environmental impact, mission completion time (e.g., fewer pauses for recharging batteries), and safety (e.g., a flying drone crashing due to poorly-managed energy consumption). Even though there is a relatively rich body of literature on energy efficiency for robotics [50], it is still unclear how practitioners are dealing with energy-related aspects in projects developed in real contexts.…”