Research Project

RecyKOOL: Maximize Waste Diversion Through Citizen Science, Data Analytics and AI Digital Twin

Research Team

Lead Researchers:

Eugene Tseng, Autonomy for Sustainability Lead
Dr. Nhut Ho, Professor in Mechanical Engineering
Dr. Bingbing Li, Professor in Manufacturing Systems Engineering

Collaborators:

WM (Waste Management)
- Mike Hammer (President – Southern California Area, Waste Management of California, Inc.)
- Kimberly Ohrt (Government Affairs)
- Jordan, Kingsbury (MRF Operations Manager)
- Stephen Moreno (Area Recycling Operations)
- Joshua, Carreon (Area Manager I)
- Mark Grady (Regional Recycling Manager)
- Kevin Vaughn (MRF Operations Manager)
- Saul Avila (MRF Operations Manager)
- Jose Figuera (Process Engineer)
Park Parthenia Apartments
- Jose Portillo (Park Parthenia Business Manager)
LAPD Devonshire PALS: Police Activity League Supporters
- Alex (Executive Director)
- Mike Lehron (Admin)
- Jose Portillo (Board Member)
Los Angeles Department of Water and Power
- Maria Sison-Roces (Utility Service Manager)

Student Team:

Troy Israel, MS Computer Engineering
John Vega, MS Software Engineering
Denver Cude, BS Computer Science
Andrew Garcia Leopold, BS Student in Computer Science
Digital Twin
- Eric Liu Xuan, PHD Candidate, Mechanical Engineering
- Nik Khandandel, MS Manufacturing Engineering
- Johnathan Aguilar, BS Computer Science
- Denver Cude, BS Computer Science
- Alex Boutselis, BS Computer Science
- David Sterin, B.S. Computer Science
- Salomon Hugo, B.S. Computer Science
- Andy Ruiz, B.S. Computer Science
- Robert Esquivel, BS Student in Mechanical Engineering
- Angel Gildo Cortes, BS Computer Science
- Mahfuz Ahmed, BS Computer Science
Citizen Science
- Yuliana Cruz, BA Student in Spanish
- Jeffrey Vasquez, BS Student in Sociology
- Robert Esquivel, BS Student in Mechanical Engineering
- Brandon Ramos, BA Student in Business Administration & Pre-Nursing
- Ciro Martinez, BS Student in Psychology
- Nico Kurthy-Bresolin, BS Student in Finance
- Jasmine Park, BS Student in Mechanical Engineering
- Jean Paul Collazo, MS Student in Data Science

Alumni Team:

Julius Maxwell, BS Student in Anthropology
Miller Alas, BS Student in Mech Engineering
Crystal Valdez, BA Student in Psychology
Jessica Reyes, BS Student in Computer Science
Digital Twin
- Cesar Aranibar David, MS in Manufacturing System Engineering
- Reza Alisamir, MS in Manufacturing System Engineering
- Bhargavkumar Gopani, MS in Manufacturing Engineering
Citizen Science
- David Gukasyan, BS Student in Business Analytics
- Lucia Miguel, BA Student in Sociology
- Tabitha Sulaiman, BS Student in Computer Science
- Jen-yu Li, MA Student in Sustainability
- Pam Porcaro, MA Student in Sustainability
- Garret Eiferman, BA Student in Psychology
- Anthony Derderian, BA Student in Political Science
- Stanislav Kazhar, BS Student in Computer Science
- Mohammed Numaan Jeelani, MS Student in Engineering Management
- Shreyas William, MS Student in Engineering Management
- Monish Ramesh, MS Student in Industrial Engineering
- Arpitha Pradeep, MS Student in Engineering Management

Funding

Funding Organization: Waste Management Inc.

SYNOPSIS

Pair citizen science with an AI-powered digital twin to increase recycling participation, reduce contamination, and improve waste diversion at a 500+ unit multifamily housing complex.
Engage residents through bilingual outreach, kitchen pail distribution, and a four-stage citizen science framework — from contributory data collection to citizen-led education and monitoring.
Simulate MRF operations using an NVIDIA Isaac Sim digital twin and train computer vision models on 9,700+ labeled images to detect contamination, equipment jams, and unsafe worker behavior in real time.
Reduce landfill disposal — currently 87% of collected tonnage — toward an optimized target of 45% trash, 20% recycling, and 35% organics.
Decrease methane emissions from landfills and support SB 1383 compliance.
Build a scalable, replicable model for urban waste management that connects upstream community behavior to downstream facility performance.
Develop workforce pathways through student internships and multidisciplinary research across engineering, computer science, sustainability, and social sciences.

Abstract

Limited awareness of proper waste sorting—distinguishing organics, recyclables, and trash—remains a persistent challenge in multifamily housing and urban communities. RecyKOOL addresses this gap by pairing community-driven data collection with AI-enabled system optimization to increase recycling participation, reduce landfill waste, and improve operational performance.

RecyKOOL is a multi-year collaboration among Waste Management (WM), the City of Los Angeles Sanitation & Environment (LASAN), the LA Local Enforcement Agency, and the CSUN Autonomy Research Center for STEAHM (ARCS). Launched in 2024 at Park Parthenia Apartments (PPA)—a 500+ unit Section 8 housing complex in Northridge, CA—the program is built on two complementary components: a citizen science initiative and an AI-powered digital twin.

The citizen science component engages residents directly in generating real-world behavioral and waste data—turning the community into active participants rather than passive. The digital twin component simulates Material Recovery Facility (MRF) operations, identifies inefficiencies, supports predictive optimization, and enables safe behavior monitoring for workers interacting with sorting equipment and waste streams. Together, these two components create a scalable, replicable model for urban waste management that connects upstream community behavior to downstream facility performance.

Motivation/Research Problem

At Park Parthenia Apartments, 87.2% of collected tonnage currently goes to trash—only 7.2% is captured through recycling and 5.6% through organics. The optimized target shifts this to 44.6% trash, 20.2% recycling, and 35.2% organics.

Apartment complexes like PPA face compounding barriers: infrastructure gaps, linguistic isolation, high residential turnover, and behavioral variability. Contamination—incorrect items in the wrong bin—reduces recovery efficiency and creates downstream challenges at the MRF, including increased downtime, equipment strain, and safety risks.

Addressing these challenges requires both behavioral intervention at the community level and system-level optimization at the facility level. Education alone is insufficient without supporting infrastructure—and infrastructure alone fails without culturally responsive, bilingual outreach.

Image Credit: Leah Marks, Yvonne Laird, Helen Trevena, Ben J. Smith, Samantha Rowbotham, “A Scoping Review of Citizen Science Approaches in Chronic Disease Prevention,” Frontiers in Public Health, 2022, DOI: 10.3389/fpubh.2022.743348.

Digital Twin Applications in Facility Management Graphic

Ideal Bin Counts at Park Parthenia Chart

Research Questions and Research Objectives

Increase recycling participation and community literacy
Improve diversion rates for recycling and organics
Reduce contamination and landfill disposal
Drive long-term behavior change through resident engagement and ambassador programs
Enhance MRF operational efficiency, safety, and uptime
Increase material recovery while reducing disposal loss
Optimize operations through data-driven decision-making
Integrate operational and business analytics for better system visibility
Enable AI-driven monitoring and predictive optimization
Support compliance with SB 1383
Strengthen overall system performance and cost efficiency

Research Methods

Step 1: Problem Identification and Awareness

Identify real-world challenges through waste assessments, bin mapping, and on-site observations. Establish baseline conditions across both community behavior and waste system performance.

Step 2: Community Engagement through Citizen Science

Engage residents through surveys, bilingual outreach, kitchen pail distribution, and participation-based activities. The citizen science framework progresses from contributory (residents track habits) → collaborative (residents flag contamination) → co-created (CSUN and WM co-design interventions with residents) → citizen-led (residents drive education and monitoring independently).

Step 3: Process Optimization with AI Digital Twin

An NVIDIA Isaac Sim digital twin simulates MRF conveyor dynamics, models both rigid and deformable waste materials, and generates synthetic training data. Utilize Digital Twin to develop a human behavior recognition system in MRF to be able to detect unsafe movements during operations in the facility. Computer vision models—YOLOv11 and Qwen2.5-VL—are trained on a custom dataset of 9,700+ labeled images across eight material categories to detect jams, contamination, and unsafe worker behaviors in real time.

Step 4: Data Analysis and Recommendations

Integrate citizen-generated data (surveys, waste audits, observations) with operational and AI-generated data to identify trends and system inefficiencies. Develop data-driven recommendations that align community behavior with optimized system performance.

Step 5: Adaptive Monitoring and Evaluation Program

Institutionalize a continuous Plan–Do–Check–Act governance cycle with monthly leadership briefings, milestone reviews, and WM executive presentations to enable iterative improvement across all program components. Implement a monitoring system of workers in pre-sort to detect unsafe behavior with the overall goal of incident prevention.

Research Results and Deliverables

AI & Digital Systems
- Implement AI digital twin (NVIDIA Isaac Sim) for predictive waste tracking
- Develop MRAG + Neo4j knowledge graph for system intelligence
- Deploy Apple Vision Pro RecyKOOL agent for immersive interaction
Tracking & Measurement
- Build tracking and measurement models to assess system performance over time
- Conduct pre/post waste composition studies
- Develop kitchen pail tracking application
- Create digital PPA bin map
Community Engagement
- Co-create scalable educational models for communities
- Launch and manage Recycling Ambassador Program
Workforce Development
- Support student internships at WM Sun Valley MRF

Commercialization Opportunities

Reduce contamination-related costs and operational inefficiencies through integrated behavioral and AI-driven solutions.
Deploy AI digital twin systems across Material Recovery Facilities for scalable optimization and predictive analytics.
Replicate the RecyKOOL framework across other WM service districts, multifamily communities, and university partners nationally.
Support SB 1383 compliance and California greenhouse gas reduction mandates through data-driven diversion strategies.

Research Timeline

Start Date: Summer 2024
End Date: TBD