Aesthetics Engineering and Computational Design
Recent advances in computational power, big data storage, and battery energy density have served as a breeding ground for autonomous driving and vehicle electrification technologies. And, the rapid advancements in these technologies have led to the emergence of a new generation of start-ups and entrants that hope to develop new products powered by autonomous electric vehicle chassis platforms, such as autonomous cargo delivery vehicle, self-driving restaurants, autonomous taxi service among others. The ability to customize chassis platforms quickly and cost-effectively therefore becomes critical in order for these start-ups to grow.
PIXBOT is a (very) easy-to-build and customizable skateboard-chassis platform built with standard aluminum profiles, powered by four electric hub motors, and a four-wheel steering system designed by PIX Moving. PIXBOT has now led the sales figures among PIX product series and has been well received by self-driving startups, R&D institutes, multi-national corporations, engineering universities and by diverse industries both in China and abroad.
We decided to develop a cargo delivery prototype on our chassis platform to showcase other potential opportunities PIXBOT could provide. This post briefly records our work and learnings absorbed during the development of the chassis shell for the cargo delivery vehicle based on PIXBOT.
Imagining The Aesthetic Look
The work began with industrial design of the cabin and the chassis platform shell. With time, the designs of both the cabin and the chassis shell kept evolving and final design (the last figure) was selected for the next steps: refinement and engineering.
Unleashing the Potential of Computational Design
Because aesthetics play a central role, we relied heavily on computational design to find out the most aesthetic design quickly. Parametric design or algorithmic design is a process that creates a relationship between a set of parameters and a geometric shape. Any changes in the parameters results a corresponding change in the design. As with any code development, creating the algorithm can be time-consuming, but once ready, it can enable the user to check a large (theoretically unlimited) number of design permutations and search for the most suitable design.
Combining parametric design codes with the mind of an industrial designer allowed us to come up with more beautiful designs much faster than just designing from scratch using traditional CAD tools.
For instance, the size of the panels was parameterized gaps to quickly figure out what gap size between them look the most appealing.
PIXBOT shell was also developed with computational design. Here is a part of the chassis shell design (Figure 3), the shell for the mid portion floor of the Chassis. The design was first parametrized, using a total of 7 parameters, which were then fed to an evolutionary algorithm. The video below shows the algorithm at play. Since the idea was to come up with a huge number of design possibilities, the parameters of the evolutionary algorithm were set randomly at first, and then fine-tuned based on the output produced. This is one way of coming up with a huge number of design possibilities in a short period of time.
PIXBOT shell is symmetric, in line with the ideal of the PIXBOT. The symmetric front and the back parts, and the side parts of the chassis were also designed computationally. The parameters here controlled the height, width, thickness, number of the vertical plates, the chamfer at the edges among other things. The final design was selected by allowing an industrial designer to play with the parameters and choose the design to his liking.
The entire code was written in rhino grasshopper, a visual programming plug in for Rhino. Grasshopper codes are highly scalable and a lot easier to use than traditional syntax-based programming languages when it comes to computational design.
As Danil Nagy explains in his post, all the functionality of Grasshopper is structured around a set of nodes which operate like functions in computer programming. Each of these nodes performs a specific operation given one or more inputs, and creates one or more outputs.
Below, a snapshot is shown to depict the size of the code for the results shown above. At the left most side in the blue box are the parameters that finally affect the design. The numeric values for these parameters can either be run by a user or can be fed to a design optimization algorithm that tends to maximize an objective function given a set of constraints. In this second case, the design becomes generative design. Figure 3 is an example of generative design. The fundamental difficulty in deploying generative design for aesthetics lies in this question — how do we measure aesthetic look? How do we tell a computer what design looks better? These questions are difficult to answer. In contrast, generative design is easy to deploy when the yardstick of measurement is well defined, such as the strength or the stiffness of a load-bearing structure. This type of generative design has been deployed commercially, in tools such as Autodesk Fusion 360.
Manufacturing Engineering
The manufacturing plans were first developed and tested iteratively by employing rapid prototyping tools. An iterative process, where improvements in the manufacturing process happen step by step, allowed us to find out faults in the manufacturing techniques quickly and find solutions. This is opposed to making a fully detailed manufacturing plan for the whole design, and hoping that the plan will work perfectly in the first step.
For instance, the installation method of the vertical plates of the chassis at the front and the back was first tried on a small scale, using 3D printed parts, then scaled to production version. Similarly, a combination of 3D printed parts, ABS plates and glue was tested in the earlier version of the design process. Simple, cheap prototypes were made quickly in order to find problems in the installation method, and then improved upon.
Prototyping helped us find many problems — that FDM based 3D printed parts were not suitable for shell from both aesthetics and engineering perspective, the suitability of the installation methods, and the type of materials, paints suitable for the shell. ABS was used as the primary material, and a Matte finish was used during painting. Acrylic was not chosen, because it cannot be painted well, also it is brittle and weaker than ABS. The problem of interference between parts were also found and solved during this process.
Learning and growth
Experimentation can thus prove valuable during the engineering of a new product, as it allows one to discover mistakes in the early stages of product development. Algorithmic design is an old technique, and it can be used for problems other than industrial design. It offers Engineers, Industrial Designers and Artists the freedom to design structures that are so often restricted by traditional design techniques. At PIX we seek to exploit this freedom and change the way cars are designed and manufactured. In the future, the customer will be able to design his own version of PIXBOT Shell.
DESIGNED AND ASSEMBLED BY PIX MOVING IN CHINA
This article is contributed by Siddharth Suhas Pawar, Mechanical Engineer at PIX Moving for more than 2 years, with a strong background in structures and composites engineering, generative design and design for additive manufacturing
