“`html
Crafting a sophisticated electronic gadget such as a delivery drone entails balancing numerous options, including the selection of motors and batteries that lower expenses while enhancing the payload capacity or the range the drone can achieve.
Deciphering that puzzle is quite challenging, but what occurs if the designers lack precise details about each battery and motor? Furthermore, the actual performance of these elements will likely be influenced by erratic factors, such as fluctuating weather conditions along the drone’s path.
Researchers at MIT have created a novel framework that assists engineers in designing intricate systems while explicitly considering such uncertainties. This framework enables them to model the performance compromises of a device with multiple interconnected components, each of which might function in unpredictable manners.
Their method captures the probability of various outcomes and compromises, providing designers with more insights than many current methodologies that, at most, can typically model only best-case and worst-case scenarios.
Ultimately, this framework could aid engineers in developing complex systems like self-driving vehicles, commercial planes, or even regional transit networks that are more robust and dependable amidst real-world unpredictability.
“In reality, the elements within a device rarely operate precisely as anticipated. If an individual has a sensor with uncertain performance, an algorithm that is also uncertain, and the design of a robot that carries uncertainties, they now have a mechanism to combine all these uncertainties to arrive at a superior design,” states Gioele Zardini, the Rudge and Nancy Allen Assistant Professor of Civil and Environmental Engineering at MIT, who is a principal investigator at the Laboratory for Information and Decision Systems (LIDS), an affiliated faculty member of the Institute for Data, Systems, and Society (IDSS), and the senior author of a paper discussing this framework.
Zardini is joined on the publication by lead author Yujun Huang, an MIT graduate student; and Marius Furter, a graduate student from the University of Zurich. The research will be showcased at the IEEE Conference on Decision and Control.
Factoring in uncertainty
The Zardini Group investigates co-design, an approach for creating systems composed of numerous interconnected elements, ranging from robots to regional transport frameworks.
The co-design framework dissects a complex issue into a sequence of boxes, with each representing an individual component, that can be combined in various configurations to optimize results or reduce expenses. This enables engineers to tackle complex issues within a practical timeframe.
In earlier research, the scientists modeled each co-design component without addressing uncertainty. For example, the performance of each sensor the designers might select for a drone was fixed.
However, engineers frequently do not possess the precise performance details for each sensor, and even if they do, it is improbable that the sensor will perfectly adhere to its specifications. Concurrently, they remain uncertain about how each sensor will perform once integrated into a complex device, or how performance will be impacted by unpredictable elements like weather.
“With our methodology, even if you are unsure of what your sensor’s specifications will be, you can still design the robot to optimize the outcome you prioritize,” remarks Furter.
To achieve this, the researchers included this concept of uncertainty into an existing framework grounded in category theory.
By employing some mathematical techniques, they streamlined the issue into a more general framework. This allows them to leverage category theory tools to resolve co-design challenges while considering a spectrum of uncertain outcomes.
By reformulating the challenge, the researchers can grasp how various design choices influence one another, even when their individual performances are uncertain.
This method is also more straightforward than many existing tools that typically demand extensive expertise in the domain. With their plug-and-play system, it is possible to rearrange the components in the system without breaching any mathematical restrictions.
And since no specific domain knowledge is needed, the framework could be utilized by a multidisciplinary team where each member contributes to designing one part of a larger system.
“Creating an entire UAV isn’t practical for just one person, but designing a component of a UAV is achievable. By providing the framework for how these components interact in a manner that considers uncertainty, we’ve simplified the evaluation of the entire UAV system’s performance,” Huang explains.
More comprehensive insights
The researchers utilized this fresh approach to select perception systems and batteries for a drone that would enhance its payload while reducing its lifetime cost and weight.
While each perception system may deliver varying detection accuracy under different weather conditions, the designer cannot determine precisely how its performance will vary. This novel system permits the designer to factor in these uncertainties when assessing the drone’s overall performance.
Moreover, unlike other methods, their framework illustrates distinct benefits of each battery technology.
For instance, their findings indicate that at lower payloads, nickel-metal hydride batteries yield the lowest anticipated lifetime cost. This insight would be unattainable without addressing uncertainty, Zardini asserts.
While another technique might only highlight the best-case and worst-case performance scenarios of lithium polymer batteries, their framework offers the user more granular information.
For example, it reveals that if the drone’s payload is 1,750 grams, there exists a 12.8 percent probability that the battery design would be unfeasible.
“Our system provides the trade-offs, enabling the user to thoughtfully consider the design,” he adds.
In the future, the researchers aim to enhance the computational efficiency of their problem-solving algorithms. They also hope to apply this approach to circumstances where a system is designed by multiple parties that are both cooperative and competitive, much like a transport network in which rail companies share the same infrastructure.
“As system complexity escalates and involves more diverse components, a formal framework for designing these systems is essential. This paper provides a means to compose large systems from modular elements, comprehend design trade-offs, and, importantly, do so while incorporating a notion of uncertainty. This creates an opportunity to formalize the design of extensive systems with learning-enabled components,” states Aaron Ames, the Bren Professor of Mechanical and Civil Engineering, Control and Dynamical Systems, and Aerospace at Caltech, who was not involved in this research.
“`