Problem Scope

In waste recycling, sorting of plastic wastes based on resin types is a critical stage. Existing workflows rely on humans for manual plastic sorting, which are not only slow and labour-intensive but are also achieving only substandard purity, resulting in degraded socioeconomical outcomes.

As pointed out by a recent publication by Royal Society of Chemistry, the lack of effective plastic waste sorting technologies is one key barrier that inhibits plastic recycling rate.

In Singapore, plastic waste (>982K tonnes generated in 2021) is among the top 4 in volume but it has the lowest recycling rates of only 6% (as compared to 99% for ferrous metals, 39% for paper and 99% for construction wastes).

Design Approach

We will develop novel multispectral self-supervised learning (MSSL) techniques to exploit the underlying structure of unlabelled data in the spectral dimension.

We will develop an automatic plastic sorting system using multi-spectral AI technologies integrated into a robot platform that can be integrated into existing waste processing pipelines, aiming to achieve a 15x increase in plastic waste sorting efficiency (with high accuracy and speed).

Potential Impact/Benefits to Target Sector 

Success would contribute to Singapore’s Zero Waste Masterplan and transformative disruptions in the waste recycling sector by (i) reducing carbon emissions due to plastic incineration, (ii) increasing the quality of materials and value derived from recycling, and (iii) enabling substantial manpower cost savings at waste processing facilities. In addition, it reduces workers’ exposure to harmful chemical substances and disease-causing pathogens from plastic waste. and reach out to your Host Institution’s research office for further details on the submission process.